- Email: [email protected]
Multiparametric Cystoscopy for Detection of Bladder Cancer Using Real-time Multispectral Imaging
EURURO-8546; No. of Pages 9 E U R O P E A N U RO L O GY X X X ( 2 019 ) X X X – X X X
available at www.sciencedirect.com journal homepage: www.europeanurology.com
Surgery in Motion
Multiparametric Cystoscopy for Detection of Bladder Cancer Using Real-time Multispectral Imaging Maximilian C. Kriegmair a,*, Jan Rother b, Bartomiej Grychtol b,c, Martin Theuring c, Manuel Ritter d, Cagatay Gu¨nes e, Maurice S. Michel a, Nikolaos C. Deliolanis b,c, Christian Bolenz e,* a
Department of Urology, University Medical Centre Mannheim, Mannheim, Germany;
b
Medical Faculty Mannheim, Heidelberg University, Mannheim,
Germany; c Project Group for Automation in Medicine and Biotechnology, Fraunhofer IPA, Mannheim, Germany; d Department of Urology, University of Bonn, Bonn, Germany; e Department of Urology, University of Ulm, Ulm, Germany
Article info
Abstract
Article history: Accepted August 15, 2019
Background: : Various imaging modalities can be used in addition to white light (WL) to improve detection of bladder cancer (BC). Objective: : To use real-time multispectral imaging (rMSI) during urethrocystoscopy to combine different imaging modalities to achieve multiparametric cystoscopy (MPC). Design, setting, and participants: : The rMSI system consisted of a camera with a spectral filter, a multi-LED light source, a microcontroller, and a computer for display and data acquisition. MSI with this system was achieved via temporal multiplexing. Surgical procedure: : MPC was performed in ten patients with a diagnosed bladder tumor. Measurements: : We gathered evidence to prove the feasibility of our approach. In addition, experienced urologists performed post-interventional evaluation of images of individual lesions. Images were independently rated in a semiquantitative manner for each modality. A statistical model was built for pairwise comparisons across modalities. Results and limitations: : Overall, 31 lesions were detected using the rMSI set-up. Histopathology revealed malignancy in 27 lesions. All lesions could be visualized simultaneously in five modalities: WL, enhanced vascular contrast (EVC), blue light fluorescence, protoporphyrin IX fluorescence, and autofluorescence. EVC and photodynamic diagnosis images were merged in real time into one MP image. Using the recorded images, two observers identified all malignant lesions via MPC, whereas the single modalities did not arouse substantial suspicion for some lesions. The MP images of malignant lesions were rated significantly more suspicious than the images from single imaging modalities. Conclusions: : We demonstrated for the first time the application of rMSI in endourology and we established MPC for detection of BC. This approach allows existing imaging modalities to be combined, and it may significantly improve the detection of bladder cancer.
Associate Editor: Alexandre Mottrie Keywords: Fluorescence imaging Multispectral Bladder cancer Transurethral resection of bladder tumor Cystoscopy Please visit www.europeanurology.com and www.urosource.com to view the accompanying video.
* Corresponding authors. Department of Urology, University Medical Center Mannheim, TheodorKutzer-Ufer 1-3, 68167 Mannheim, Germany. Tel. +49 621 3831259, Fax: +49 621 3832184. E-mail address: [email protected] (M.C. Kriegmair). Department of Urology, University of Ulm, Albert-Einstein-Allee 23, 89081 Ulm, Germany. Tel. +49 731 50058000; Fax: +49 731 50058002. E-mail address: [email protected] (C. Bolenz).
https://doi.org/10.1016/j.eururo.2019.08.024 0302-2838/© 2019 European Association of Urology. Published by Elsevier B.V. All rights reserved.
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 2
E U RO P E A N U R O L O GY X X X ( 2 019 ) X X X – X X X
Patient summary: : Real-time multispectral imaging was successfully used to combine different imaging aids for more comprehensive illustration of bladder tumors for surgeons. In the future, this technique may allow better detection of bladder tumors and more complete endoscopic resection in cases of cancer. © 2019 European Association of Urology. Published by Elsevier B.V. All rights reserved.
1.
Introduction
Bladder cancer (BC) is among the most common cancers worldwide [1]. High recurrence and progression rates are major unsolved problems and place high burdens on health care systems [2]. Almost 70% of BCs are non–muscleinvasive at diagnosis and can be managed with repeated endoscopic resection and instillation therapies. BC is diagnosed and pathologically staged via endoscopy and transurethral resection of the bladder (TUR-B). Patients require long-term cystoscopy follow-up to detect recurrences [2,3]. White light (WL) cystoscopy is the standard procedure for identifying suspicious lesions of the bladder. It has excellent sensitivity for detection of papillary lesions, and current technical developments such as digital endoscopes with high-definition technology have further refined WL imaging [4,5]. Nonetheless, flat cancerous tissue (carcinoma in situ, CIS) or very small tumors may remain undetected in WL, even when modern endoscopes are used [6]. However, detection of CIS and resection of all tumors are essential for risk stratification and clinical decision-making, and they prevent recurrence effectively [7]. To improve detection rates, WL imaging can be combined with additional imaging modalities. Photodynamic diagnosis (PDD) involves preoperative bladder instillation of a substrate of heme metabolism. Subsequently, protoporphyrin IX (PpIX) accumulates in malignant cells and emits red fluorescence under excitation with blue light [8]. PDD has shown a clear advantage over WL in detecting flat lesions. Up to 40% of all CIS can only be identified under PDD according to a recent meta-analysis [6,9]. Narrow-band imaging (NBI) uses light of defined wavelengths that are strongly absorbed by the hemoglobin in vessels of the bladder mucosa. This leads to enhanced vascular contrast (EVC) that helps to identify potentially malignant lesions with increased or abnormal vasculature. As for PDD, studies have shown that NBI has clear value in increasing the rate of BC detection [10]. We recently showed that tissue autofluorescence (AF) can also assist in distinguishing normal and malignant bladder-wall tissue in wide-field endoscopy [11]. Further imaging modalities such as digital contrast enhancement and near-infrared imaging are currently under evaluation [12]. All the various adjunct imaging modalities have individual drawbacks and limitations. Modalities that require specific illumination (PDD, NBI, and AF), for instance, can only be visualized separately, but not in parallel with or overlaid with WL. Hence, repetitive switching during TUR-B is required. Moreover, existing endoscopic systems cannot combine multiple imaging modalities [13]. Real-time multispectral imaging (rMSI) allows separate visualization
of multiple spectral components and thus can extract information not visible in WL images [14]. The aim of our study was to adapt an rMSI device for urethrocystoscopy to allow visualization and combination of multiple adjunct imaging modalities (PDD, EVC, AF) with WL imaging. This methodology may allow multiparametric cystoscopy (MPC) for detection of BC for the first time. 2.
Patients and methods
2.1.
Imaging setup
The general setup for rMSI consists of a camera unit, a light source, and a computer with a microcontroller board for control of the camera and light source (Fig. 1). A color scientific complementary metal-oxide-semiconductor camera (PCO AG, Kelheim, Germany) is used for imaging. The endoscope is mounted on the camera using a C-mount adapter (R. Wolf, Knittlingen, Germany). An optical multibandpass filter (Chroma Technology, Bellows Falls, VT, USA) is mounted on the C-mount adapter using a custom 3Dprinted mount and two-phase dental glue (Picodent Dental Produktions und Vertriebs, Wipperfürth, Germany). A modular LED light source (Omicron-laserage Laserprodukte, Rodgau, Germany) with peak wavelengths distributed over the visible range is used. The LEDs are individually filtered to match the bandpass properties of the filter in front of the camera sensor. The camera and light source are controlled via a microcontroller board interfaced with bespoke software running on a high-end personal computer, which is also used for basic image processing and displaying the video stream on the screen in real time during cystoscopy. For cystoscopy, sterilized PDD endoscopes (30 optics, Karl Storz, Tuttlingen, Germany) and sterilized liquid light guides (Karl Storz) are used. MSI is achieved via temporal multiplexing of WL, EVC, and PDD illumination, providing a video stream with a frame rate of approximately 20 Hz. 2.2.
Image processing
Frames of color imaging (WL) and total fluorescence (PDD) are white balanced, and the EVC frames are mapped to different RGB channels of the output image. In postprocessing (but not the live view), the contrast of WL and EVC frames is also increased by applying contrast-limited adaptive histogram equalization, and the saturation of the ECV and PDD images is also increased. The MP image is obtained via digital fusion of the EVC and PDD images (Fig. 2A). This is possible in real time, allowing for real-time MP imaging. For standardization, the MP images reviewed by urologists (see Section 2.4) were
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 E U RO P E A N U RO L O GY X X X ( 2 019 ) X X X – X X X
3
Fig. 1 – Schematic diagram of the real-time multispectral imaging system. The system consists of a scientific complementary metal-oxidesemiconductor camera with a C-mount, a video adapter with a glued optical filter, and a light source. The camera and light source are controlled via a microcontroller. A personal computer (PC) records the data and displays the videos to the surgeon.
generated after the intervention. The PpIX fluorescence (PpIX-F) and AF modalities are derived from the PDD modality via linear unmixing with an empirically determined unmixing matrix (Fig. 2B). Thus, the resulting component images are monochrome and are colored afterwards for better visualization.
that WL, EVC, PDD (and thus PpIX and AF, but not unmixed), and the merged MP image were available during cystoscopy. Images of all rMSI modalities analyzed were generated after interventions from the raw data using a standardized procedure as described above. 2.4.
2.3.
Image analysis and statistical evaluation
Patients and surgical procedure
Overall, ten patients scheduled for TUR of a cystoscopically diagnosed bladder tumor were enrolled. All patients provided written informed consent for participation in this pilot study, which was approved by the local ethics committee (approval number 2015-616N-MA). All patients received a bladder instillation of 85 mg of hexaminolevulinate (Hexvix, Ipsen Pharma, Boulogne-Billancourt, France) 1 h before the procedure. Surgery was performed under general or spinal anesthesia. MSI was used to scan the entire bladder for suspicious lesions. Video data were recorded synchronously. After MSI, conventional endoscopy using the D-Light System with a Tricam SL II camera and a Ch26 continuous-flow shaft with 30 lenses (Karl Storz) was performed. All suspicious lesions identified with the standard instrument were biopsied or resected for a standard histological examination. For bipolar resection, an Autocon II 400 SCB system (Karl Storz) was used. Initially, WL, EVC, and PDD were available in real time during MSI for all patients. During the course of the study we were able to further integrate real-time MP imaging, so
After surgery, the video sequences were reviewed by the operating surgeon to identify frames showing the resected lesions. Images of those lesions in all six modalities (WL, EVC, PDD, PpIX-F, AF, and MP) were independently reviewed in random order by two urologists and scored on a fourpoint Likert scale, with 0 denoting “not suspicious” and 3, “clearly suspicious”. The scores for the 27 malignant lesions (n = 324 scores) were further analyzed using STATA (StataCorp LLC, College Station, TX, USA). To tease out the effect of the modality under which a lesion was observed on the degree of suspicion it aroused, a multilevel mixed-effects ordered logistic regression (meologit command) was built with fixed effects for modality, images nested in lesions and controls for a patient, reviewer (surgeon), and whether the image contained a flat lesion. From the resultant model, the predicted probabilities of each score for each modality were estimated with 95% confidence intervals and compared pairwise for each score. The choice of the statistical model was motivated by the ordinal character of the dependent variable (score), multiple
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 4
E U RO P E A N U R O L O GY X X X ( 2 019 ) X X X – X X X
Fig. 2 – (A) Merger of enhanced vascular contrast (EVC) luminescence-based contrast and photodynamic diagnosis (PDD) fluorescence-based contrast into one multiparametric (MP) image. (B) Unmixing of full PDD fluorescence image into protoporphyrin IX fluorescence (PpIX-F) and autofluorescence (AF) components.
Table 1 – Patient and tumor characteristics for the pilot study population. Number
Sex
Age (yr)
Recurrence
Tumor size (cm)
Lesions (n)
Histology
1 2 3 4 5 6 7 8 9 10
M M M M M M F M M M
70 74 61 79 73 73 64 55 81 74
No Yes Yes No No Yes No Yes Yes No
<3 <3 <3 >3 >3 >3 >3 >3 >3 <3
2 3 4 1 2 10 1 1 4 2
Benign UCC UCC UCC UCC UCC SCC UCC UCC UCC
TMN
Grade
– pTa pTa pTa pT1 pTa >pT2 pTa pTa + Tis pTa
– LG LG LG HG LG HG HG LG + HG LG
M = male; F = female; UCC = urothelial cell carcinoma; SCC squamous cell carcinoma; HG = high grade; LG = low grade.
observations of the same lesion in different modalities, and possible patient- and reviewer-specific influences. The likelihood ratio test against the simpler ordinary logistic regression model was significant (p = 0.0055), indicating that there is sufficient variance between lesions to warrant a multilevel model. Robustness checks against the simpler model with clustered standard errors or fixed effects for lesion did not produce substantially different results. In addition, keeping in mind that all images contained a malignant lesion, we transformed the scores into a dichotomous variable, defining scores of 0 and 1 as failure, and scores of 2 and 3 as success for each modality in arousing surgeon suspicion. We coded failure as 1 and success as 0. A multilevel, mixed-effects logistic regression model was constructed with fixed effects for modality, images nested in lesions and controls for the reviewer, and flat lesion. This allowed us to estimate the predicted failure
rate for each modality with 95% confidence intervals and to obtain corresponding p values. The MP modality was dropped from the model because no case of failure was present in the data.
3.
Results
rMSI was successfully performed in ten patients (nine male, one female) with diagnosed bladder tumors referred for TUR-BT. Patient characteristics are displayed in Table 1. Half of the patients had a history of previous BC. Malignant pathology was found in nine patients. The median number of lesions per patient was two (range 1–10). The median interventional time was 31 (11–120) min and no intraoperative or postoperative complications occurred. Overall, 31 lesions were assessed via rMSI. The lesion characteristics
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 5
E U RO P E A N U RO L O GY X X X ( 2 019 ) X X X – X X X
Table 2 – Macroscopic and pathological description of the 30 lesions identified and semiquantitative rating results from two urologistsa. Patient
Lesion
Histology
TMN
1
1 2 1 2 3 1 2 3 4 1 1 2 1 2 3 4 5 6 7 8 9 10 1 1 1 2 3 4 1 2 3
BI BI UCC UCC UCC UCC UCC UCC BI UCC UCC UCC UCC UCC UCC UCC UCC UCC UCC UCC UCC UCC sUCC UCC UCC UCC BI UCC UCC UCC UCC
– – pTa pTa pTa pTa pTa pTa
LG LG LG LG LG LG
pTa pT1 pTa pTa pTa pTa pTa pTa pTa pTa pTa pTa pTa >pT2 pTa Tis pTa
LG HG HG LG LG LG LG LG LG LG LG LG LG HG HG HG LG
Tis pTa pTa pTa
HG LG LG LG
2
3
4 5 6
7 8 9
10
Grade
Location
4 4 5 5 5 7 5 5 4 1, 7 7 7 4 7, 2 7 7 8 6, 8 1 2, 4 4 8 8 8, 2 1 1 7 9 4 2 2
Architecture
WL
Flat sessile Flat Papillary Papillary Papillary Sessile Sessile Microsessile Microsessile Flat papillary Exophytic sessile Exophytic sessile Papillary Flat papillary Papillary Small papillary Papillary Flat papillary Small papillary Flat Flat Small sessile Large sessile Flat papillary Flat papillary Flat papillary Flat Flat Papillary Papillary Papillary
U1 0 1 3 3 3 2 2 0 3 3 3 3 3 3 1 3 3 3 1 1 2 3 2 3 3 1 0 3 3 3
U2 1 2 3 2 2 3 3 0 0 3 2 3 3 3 2 1 3 3 3 1 0 2 2 1 2 3 2 1 2 3 2
EVC U1 1 2 3 3 3 2 2 0 1 3 3 3 3 3 3 2 3 3 3 1 2 3 3 2 2 3 1 1 3 3 2
U2 1 2 2 3 3 1 2 0 1 3 2 2 3 3 3 2 3 2 3 2 1 3 2 2 2 3 2 1 3 3 2
PDD U1 1 0 2 2 3 2 2 2 0 3 3 3 3 3 3 3 2 2 3 3 2 3 2 1 3 2 2 2 3 3 3
U2 1 1 3 2 3 2 1 2 1 3 3 3 3 3 3 3 2 2 3 3 2 3 1 1 3 2 2 3 2 2 2
PpIX-F U1 0 0 2 2 2 2 1 2 0 3 3 3 3 3 3 2 2 2 3 3 1 3 1 1 3 2 0 2 3 3 3
U2 0 0 3 2 2 2 2 2 0 3 3 3 3 3 3 2 3 2 3 3 1 3 1 1 3 2 0 1 2 3 3
AF U1 0 2 3 1 2 2 2 2 0 3 1 1 1 3 3 3 2 1 3 3 2 3 1 1 2 1 1 1 3 2 3
MP U2 0 1 3 1 1 2 1 2 0 3 2 2 3 3 2 3 2 1 3 2 2 2 2 2 1 1 2 2 2 2 2
U1 1 2 3 3 3 3 2 2 0 2 3 3 3 3 3 3 3 3 3 3 2 3 3 3 3 3 1 3 3 3 3
U2 1 2 3 3 3 3 3 2 1 2 3 3 3 3 3 3 3 3 3 3 2 3 3 2 3 3 2 3 3 3 3
BI = benign inflammation; UCC = urothelial cell carcinoma; sUCC = squamous cell urothelial carcinoma; HG = high grade; LG = low grade; WL = white light; EVC = enhanced vascular contrast; PDD = photodynamic diagnosis; PpIX-F = protoporphyrin IX fluorescence; AF = autofluorescence; MP = multiparametric. a The two physicians (U1 and U2) scored lesion suspicion on the different imaging modalities on a scale from 0 to 3.
are summarized in Table 2. Histopathology revealed a malignant tumor in 27 lesions, of which the majority were pTa low grade (n = 22). A high-grade tumor was identified in five lesions, of which two were CIS and one was a muscleinvasive, squamous-like tumor. rMSI visualized the bladder mucosa using six different modalities: WL, EVC, PDD, PpIX-F, AF, and MP. No technical difficulties occurred. All lesions identified with the rMSI system were judged suspicious on standard PDD, and therefore biopsies were taken. Furthermore, standard PDD did not identify any additional lesions previously not detected with rMSI. Fig. 3 shows different lesions in all modalities. Lesions 1 and 2 are typical papillary bladder tumors that are clearly visible in all modalities. Both tumors have a positive PpIX-F and a negative AF signal. Lesion 3 is a flat tumor at the bladder trigonum. Under WL, the lesion is hardly demarcated from the surrounding mucosa. EVC mode allows identification due to differences between the tumor and the adjacent mucosa. In PDD mode, the tumor is clearly highlighted. Under separate PpIX-F and AF displays, the putatively malignant tissue exhibits strong PpIX-F and weak AF, whereas normal mucosa exhibits strong AF signals. In the MP image, the specific characteristics of the single modalities are overlaid on the WL image. Lesion 4 comprises very small papillary lesions on the back wall of the bladder. The different modalities again contribute to
clear demarcation of the malignant lesions, and they are combined in the MP image. Lesion 5 is CIS on the left bladder wall. Positive PpIX-F and weak AF arouse suspicion of a malignant process. Images of all 27 malignant lesions in each modality were rated by two urologists on a Likert scale from 0 (not suspicious) to 3 (clearly suspicious). The results are reported in Table 2 and presented graphically in Fig. 4. MP was the only modality for which all images were scored as 2 or 3; other modalities also received lower ratings. In addition, the fraction of images scoring 3 was highest for the MP modality (>0.8). Multilevel ordinal logistic regression for these ratings revealed a significant positive effect of the MP modality on score (p < 0.001) and a negative effect of the AF modality (0.046). Images showing flat lesions in general received lower scores (p < 0.001). The predicted probability of completely missing a lesion (score 0 on the Likert scale) was significantly greater than zero for all modalities other than MP (p = 0.128; Fig. 4). In particular, the predicted probability of WL images being scored 0 was 0.0234 (p = 0.042), which was significantly higher (p = 0.045) than for MP (0.002; p = 0.128). The probability of an MP image being scored 1 was 0.016, which was significantly lower than all other modalities (all p < 0.001) and in particular PDD (0.100), EVC (0.115), and WL (0.138). MP images had the highest predicted probability of
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 6
E U RO P E A N U R O L O GY X X X ( 2 019 ) X X X – X X X
Fig. 3 – Examples of lesion images in all modalities. Lesions 1 and 2 are typical papillary tumors and are clearly visible in all modalities. Both tumors show strong protoporphyrin IX fluorescence (PpIX-F) and negative autofluorescence (AF). Final histology was pTa low grade. Lesion 3 is a small and flat tumor (pTa low grade) at the bladder trigonum and is barely detectable with white light (WL). Its abnormal vasculature is demarcated on enhanced vascular contrast (EVC) mode. It also shows positive PpIX-F and low AF. All features are overlaid in the multiparametric (MP) mode. Lesion 4 consists of very small papillary lesions (pTa low grade). Comparable to lesion 3, the single modalities contribute to clearer identification of the tumors. Lesion 5 is a flat lesion with the final pathology of carcinoma in situ. It has an abnormal vasculature highlighted by EVC with positive PpIX-F and negative AF. The MP mode clearly demarcates the lesions and their full extent.
receiving a score of 3 (clearly suspicious). The second highest was PDD at 0.567, with WL at 0.474. The probabilities of each modality failing to arouse substantial suspicion as predicted by the logistic regression model for the dichotomous variable are presented in Fig. 5. With no failures, MP was excluded from this analysis. Among the other modalities, all had a probability of failure of significantly greater than zero (p < 0.001 for WL and AF, others up to p = 0.036). Only PDD had a significantly lower probability of failure than WL (0.075 vs 0.200; p = 0.039). Failure was more likely for images containing flat lesions (p = 0.006). There was no statistically significant difference between the two reviewers (p = 0.852). 4.
Discussion
WL endoscopy currently remains the imaging gold standard in endourology. It offers high sensitivity and specificity for most papillary bladder tumors [4,5]. However, it has major limitations in the detection of flat and small lesions, up to 50% of which are missed under WL [6]. Additional imaging modalities such as PDD and NBI can add information and support urologists in identifying malignant urothelial lesions [13]. Various studies have shown a benefit of PDD and NBI in increasing detection
rates. In a meta-analysis, Burger et al [9] found that 40% of flat lesions can only be detected when using PDD. Eventually, use of PDD during resection might even contribute to an increase in recurrence-free survival [6]. Similar data exist for NBI [10]. Using AF emitted by endogenous fluorophores in the mucosal layer, epithelial cancer could be detected in different endoscopic fields (eg, bronchoscopy and colonoscopy) [15,16]. We recently demonstrated that AF may also improve the rate of BC detection [11]. Unfortunately, all these additional modalities have individual drawbacks and they can only be applied separately. So far, a combination or even simultaneous application has not been possible. We present the first human application of MP endoscopy in urology. Using an rMSI camera and software system, we could visualize the bladder mucosa in up to five different modalities simultaneously. Fusion of specific information from these modalities in one endoscopic image was feasible. This technical approach allowed MPC with astounding and diverse potential. Beside demonstrating feasibility, our preliminary data suggest a better detection rate, especially for small and flat malignant lesions of the bladder wall. In our pilot study, all malignant lesions were rated as suspicious using the MP modality, whereas observers rated some malignant lesions as not or only a little suspicious
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 E U RO P E A N U RO L O GY X X X ( 2 019 ) X X X – X X X
7
Fig. 4 – Fraction and predicted probability with 95% confidence interval of each score for each modality. Fractions were calculated directly from histograms of the scores assigned by reviewers to images containing malignant lesions. Predicted probabilities are from the multilevel ordinal logistic regression model after estimation. MP = multiparametric; WL = white light; EVC = enhanced vascular contrast; PDD = photodynamic diagnosis; PpIX = protoporphyrin IX fluorescence; AF = autofluorescence.
when using the single modalities. We thus combined the individual benefits of each modality in a merged image that compensated for the limitations of the individual modalities. Some malignant lesions may not appear suspicious in some modalities. Data from a recent study suggest that more malignant flat lesions might be detected when PDD and NBI are used subsequently [17]. PDD has a failure rate in routine clinical practice, and some malignant tumors show no or little fluorescence. In an MP approach, this could be compensated by other modalities such as EVC and AF. Moreover, PDD/PpIX-F, EVC, and AF are excellent complements to each other: PDD/PpIX-F gives information on the metabolic activity of the tissue by highlighting mucosa with high uptake of 5-aminolevulinic acid [18]. EVC draws the observer’s eye to highly vascularized and thus potentially malignant tissue [10]. AF provides information on thickening and abnormal composition of the epithelial layer. We previously found that AF has a high negative predicitve value, suggesting that a strong AF signal reliably indicates a normal healthy mucosa [11]. Consequently, AF is a suitable supplement to PDD/PpIX-F and EVC, which predominantly highlight potentially malignant tissue. This underlines the unique advantages of the MP approach. So far, some providers have introduced endoscopic platforms that
present different images to the urologist, such as the SPIES system (Karl Storz), which uses digital image processing and contrast enhancement to highlight different aspects of the image (eg, vasculature, depth, and illumination) [12]. However, this approach is limited to data acquired only from WL images, and no additional information on metabolism or epithelial composition is displayed for the physician. By contrast, the objective of MP imaging is to combine modalities with different metabolic and anatomic implications in a true MP approach. MP imaging technology enhances WL cystoscopy and offers high diagnostic potential not only for better BC detection but also for more complete TUR. Most surgeons currently use additional imaging modalities to detect putatively malignant lesions, but they prefer to resect under WL. This carries the risk of imprecise tumor resection, especially at the edges of flat and diffusely growing lesions. Incomplete resection is believed to be at least partly responsible for a tumor detection rate of up to 40% during secondary TUR-BT [19]. By providing WL images and additional imaging information (eg, fluorescence) during the resection in overlay and in parallel, the MP cystoscopy technology might improve resection and reduce the remaining cancerous tissue. The proof of the assumption
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 8
E U RO P E A N U R O L O GY X X X ( 2 019 ) X X X – X X X
Fig. 5 – Predicted probability of failure for each modality with 95% confidence interval obtained from multilevel mixed-effects logistic regression on the dichotomous variable after estimation. MP = multiparametric; WL = white light; EVC = enhanced vascular contrast; PDD = photodynamic diagnosis; PpIX = protoporphyrin IX fluorescence; AF = autofluorescence.
requires further studies. However, there might be a predisposition to a higher rate of false-positive findings, especially during MP TUR-BT, since abnormal findings in any modality might induce a decision to biopsy. In the future, training and experience in MP cystoscopy as well as computer-aided systems may support surgeons in avoiding unnecessary biopsies. Future MPC perspectives could include the application of fluorescence-labeled targeted antibodies to specifically label BC. A recent ex vivo study used CD47 antibodies to mark BC for blue-light endoscopy and confocal endomicroscopy in ex vivo radical cystectomy specimens [20]. Higher sensitivity for detection of malignant lesions and molecular marking may also help to identify variants of BC that are diagnostically challenging. Moreover, specificity might increase, which would improve diagnostic accuracy for CIS or recurrent BC following prior intravesical immunotherapy with bacillus Calmette-Guérin [20]. MP endoscopy in theory allows combination with antibody-based detection and could thus potentially further improve detection rates. Furthermore, the inclusion of other potential molecular markers such as EGFR and PSCA—with the use of differentially labeled antibodies—could be useful for specifying findings, in line with the idea of molecular endoscopy [21,22]. In the current MP imaging implementation, differentiation (and thus combination of up to three different fluorophores) is feasible, which opens the door to a diverse molecular approach. Molecular endoscopy needs a multistep preclinical evaluation process, for which an in vivo model has been established by our group [23]. Our study presents initial clinical data on a new imaging device. It is limited by its proof-of-concept design, with a relatively small number of patients and lesions assessed. Although we have demonstrated the feasibility of MP endoscopy for detection of BC, the added value of MPC in clinical routine needs to be validated. Moreover, resection was performed using conventional endoscopic equipment (WL and PDD) following MPC, and thus the value of MPC during tumor resection needs to be examined further. Ultimately, MPC needs to be compared to existing state-ofthe-art endoscopic imaging modalities, preferably in a head-to-head study.
5.
Conclusions
In this pilot study, we enhanced cystoscopy by implementing rMSI in endourology and by combining different imaging modalities. This allowed us to perform MPC for the first time for the diagnosis of malignant bladder lesions. Up to five different modalities (WL, EVC, PDD, PpIX-F, and AF) were displayed simultaneously and combined into an MP image. Our preliminary results suggest better detection of bladder tumors using MPC. This technology may further complement fluorescent signals, for example, by using fluorescence-labeled specific antibodies for molecular endoscopy. Author contributions: Maximilian C. Kriegmair had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Kriegmair, Bolenz, Deliolanis. Acquisition of data: Kriegmair, Rother, Theuring. Analysis and interpretation of data: Kriegmair, Grychtol, Bolenz. Drafting of the manuscript: Kriegmair, Rother, Grychtol. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: Grychtol. Obtaining funding: Bolenz, Deliolanis. Administrative, technical, or material support: Rother, Theuring, Grychtol, Deliolanis, Günes. Supervision: Bolenz, Deliolanis. Other (video): Rother, Kriegmair. Financial disclosures: Maximilian C. Kriegmair certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Nikolaos C. Deliolanis is a co-inventor for multispectral imaging patent applications. The remaining authors have nothing to disclose.
Funding/Support and role of the sponsor: This work was supported by funding from the German Ministry for Education and Research (GO-Bio Project 031B0219). The sponsor played no direct role in the study.
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 E U RO P E A N U RO L O GY X X X ( 2 019 ) X X X – X X X
Appendix A. Supplementary data
9
[11] Kriegmair MC, Honeck P, Theuring M, Bolenz C, Ritter M. Wide-field autofluorescence-guided TUR-B for the detection of bladder cancer:
The Surgery in Motion video accompanying this article can be found in the online version at doi:https://doi.org/10. 1016/j.eururo.2019.08.024.
a pilot study. World J Urol 2018;36:745–51. [12] Kamphuis GM, de Bruin DM, Brandt MJ, et al. Comparing image perception of bladder tumors in four different Storz Professional Image Enhancement System modalities using the iSPIES app. J Endourol 2016;30:602–8.
References
[13] Lerner SP, Goh A. Novel endoscopic diagnosis for bladder cancer. Cancer 2015;121:169–78.
[1] Antoni S, Ferlay J, Soerjomataram I, Znaor A, Jemal A, Bray F. Bladder
[14] Dimitriadis N, Grychtol B, Theuring M, Behr T, Sippel C, Deliolanis
cancer incidence and mortality: a global overview and recent
NC. Spectral and temporal multiplexing for multispectral fluores-
trends. Eur Urol 2017;71:96–108. [2] Wong MCS, Fung FDH, Leung C, Cheung WWL, Goggins WB, Ng CF. The global epidemiology of bladder cancer: a joinpoint regression
cence and reflectance imaging using two color sensors. Opt Express 2017;25:12812–29. [15] He Q, Wang Q, Wu Q, Feng J, Cao J, Chen BY. Value of autofluores-
analysis of its incidence and mortality trends and projection. Sci Rep
cence imaging videobronchoscopy in detecting lung cancers and
2018;8:1129.
precancerous lesions: a review. Respir Care 2013;58:2150–9.
[3] Burger M, Catto JW, Dalbagni G, et al. Epidemiology and risk factors of urothelial bladder cancer. Eur Urol 2013;63:234–41.
[16] Li BH, Xie SS. Autofluorescence excitation-emission matrices for diagnosis of colonic cancer. World J Gastroenterol 2005;11:3931–4.
[4] Borin JF, Abdelshehid CS, Clayman RV. Comparison of resolution,
[17] Drejer D, Beji S, Oezeke R, et al. Comparison of white light, photo-
contrast, and color differentiation among fiberoptic and digital
dynamic diagnosis, and narrow-band imaging in detection of car-
flexible cystoscopes. J Endourol 2006;20:54–8. [5] Cina SJ, Epstein JI, Endrizzi JM, Harmon WJ, Seay TM, Schoenberg MP. Correlation of cystoscopic impression with histologic diagnosis of biopsy specimens of the bladder. Hum Pathol 2001;32:630–7. [6] Rink M, Babjuk M, Catto JW, et al. Hexyl aminolevulinate-guided fluorescence cystoscopy in the diagnosis and follow-up of patients with non–muscle-invasive bladder cancer: a critical review of the current literature. Eur Urol 2013;64:624–38. [7] Soukup V., Capoun O., Cohen D., et al. Risk stratification tools and
cinoma in situ or flat dysplasia at transurethral resection of the bladder: the DaBlaCa-8 study. Urology 2017;102:138–42. [18] Nakai Y, Tatsumi Y, Miyake M, et al. Expression of ferrochelatase has a strong correlation in protoporphyrin IX accumulation with photodynamic detection of bladder cancer. Photodiagn Photodyn Ther 2016;13:225–32. [19] Schwaibold HE, Sivalingam S, May F, Hartung R. The value of a second transurethral resection for T1 bladder cancer. BJU Int 2006;97:1199–201.
prognostic models in non-muscle-invasive bladder cancer: a critical
[20] Pan Y, Volkmer JP, Mach KE, et al. Endoscopic molecular imaging of
assessment from the European Association of Urology Non–Muscle-
human bladder cancer using a CD47 antibody. Sci Transl Med
Invasive Bladder Cancer Guidelines Panel. Eur Urol Focus. In press. https://doi.org/10.1016/j.euf.2018.11.005.
2014;6:260ra148. [21] Rebouissou S, Bernard-Pierrot I, de Reynies A, et al. EGFR as a
[8] Kriegmair M, Baumgartner R, Knuchel R, Stepp H, Hofstadter F,
potential therapeutic target for a subset of muscle-invasive bladder
Hofstetter A. Detection of early bladder cancer by 5-aminolevulinic
cancers presenting a basal-like phenotype. Sci Transl Med
acid induced porphyrin fluorescence. J Urol 1996;155:105–9.
2014;6:244ra91.
[9] Burger M, Grossman HB, Droller M, et al. Photodynamic diagnosis of
[22] Yuan R, Rao T, Cheng F, et al. Quantum dot-based fluorescent probes
non-muscle-invasive bladder cancer with hexaminolevulinate cys-
for targeted imaging of the EJ human bladder urothelial cancer cell
toscopy: a meta-analysis of detection and recurrence based on raw data. Eur Urol 2013;64:846–54.
line. Exp Ther Med 2018;16:4779–83. [23] Günes C, Meessen S, Rother J, et al. Development of a rodent model
[10] Zheng C, Lv Y, Zhong Q, Wang R, Jiang Q. Narrow band imaging
for preclinical evaluation of multiple contrast agents and real-time
diagnosis of bladder cancer: systematic review and meta-analysis.
multispectral imaging in bladder cancer. Eur Urol Suppl 2019;18:
BJU Int 2012;110:E680–7.
e1955.
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
available at www.sciencedirect.com journal homepage: www.europeanurology.com
Surgery in Motion
Multiparametric Cystoscopy for Detection of Bladder Cancer Using Real-time Multispectral Imaging Maximilian C. Kriegmair a,*, Jan Rother b, Bartomiej Grychtol b,c, Martin Theuring c, Manuel Ritter d, Cagatay Gu¨nes e, Maurice S. Michel a, Nikolaos C. Deliolanis b,c, Christian Bolenz e,* a
Department of Urology, University Medical Centre Mannheim, Mannheim, Germany;
b
Medical Faculty Mannheim, Heidelberg University, Mannheim,
Germany; c Project Group for Automation in Medicine and Biotechnology, Fraunhofer IPA, Mannheim, Germany; d Department of Urology, University of Bonn, Bonn, Germany; e Department of Urology, University of Ulm, Ulm, Germany
Article info
Abstract
Article history: Accepted August 15, 2019
Background: : Various imaging modalities can be used in addition to white light (WL) to improve detection of bladder cancer (BC). Objective: : To use real-time multispectral imaging (rMSI) during urethrocystoscopy to combine different imaging modalities to achieve multiparametric cystoscopy (MPC). Design, setting, and participants: : The rMSI system consisted of a camera with a spectral filter, a multi-LED light source, a microcontroller, and a computer for display and data acquisition. MSI with this system was achieved via temporal multiplexing. Surgical procedure: : MPC was performed in ten patients with a diagnosed bladder tumor. Measurements: : We gathered evidence to prove the feasibility of our approach. In addition, experienced urologists performed post-interventional evaluation of images of individual lesions. Images were independently rated in a semiquantitative manner for each modality. A statistical model was built for pairwise comparisons across modalities. Results and limitations: : Overall, 31 lesions were detected using the rMSI set-up. Histopathology revealed malignancy in 27 lesions. All lesions could be visualized simultaneously in five modalities: WL, enhanced vascular contrast (EVC), blue light fluorescence, protoporphyrin IX fluorescence, and autofluorescence. EVC and photodynamic diagnosis images were merged in real time into one MP image. Using the recorded images, two observers identified all malignant lesions via MPC, whereas the single modalities did not arouse substantial suspicion for some lesions. The MP images of malignant lesions were rated significantly more suspicious than the images from single imaging modalities. Conclusions: : We demonstrated for the first time the application of rMSI in endourology and we established MPC for detection of BC. This approach allows existing imaging modalities to be combined, and it may significantly improve the detection of bladder cancer.
Associate Editor: Alexandre Mottrie Keywords: Fluorescence imaging Multispectral Bladder cancer Transurethral resection of bladder tumor Cystoscopy Please visit www.europeanurology.com and www.urosource.com to view the accompanying video.
* Corresponding authors. Department of Urology, University Medical Center Mannheim, TheodorKutzer-Ufer 1-3, 68167 Mannheim, Germany. Tel. +49 621 3831259, Fax: +49 621 3832184. E-mail address: [email protected] (M.C. Kriegmair). Department of Urology, University of Ulm, Albert-Einstein-Allee 23, 89081 Ulm, Germany. Tel. +49 731 50058000; Fax: +49 731 50058002. E-mail address: [email protected] (C. Bolenz).
https://doi.org/10.1016/j.eururo.2019.08.024 0302-2838/© 2019 European Association of Urology. Published by Elsevier B.V. All rights reserved.
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 2
E U RO P E A N U R O L O GY X X X ( 2 019 ) X X X – X X X
Patient summary: : Real-time multispectral imaging was successfully used to combine different imaging aids for more comprehensive illustration of bladder tumors for surgeons. In the future, this technique may allow better detection of bladder tumors and more complete endoscopic resection in cases of cancer. © 2019 European Association of Urology. Published by Elsevier B.V. All rights reserved.
1.
Introduction
Bladder cancer (BC) is among the most common cancers worldwide [1]. High recurrence and progression rates are major unsolved problems and place high burdens on health care systems [2]. Almost 70% of BCs are non–muscleinvasive at diagnosis and can be managed with repeated endoscopic resection and instillation therapies. BC is diagnosed and pathologically staged via endoscopy and transurethral resection of the bladder (TUR-B). Patients require long-term cystoscopy follow-up to detect recurrences [2,3]. White light (WL) cystoscopy is the standard procedure for identifying suspicious lesions of the bladder. It has excellent sensitivity for detection of papillary lesions, and current technical developments such as digital endoscopes with high-definition technology have further refined WL imaging [4,5]. Nonetheless, flat cancerous tissue (carcinoma in situ, CIS) or very small tumors may remain undetected in WL, even when modern endoscopes are used [6]. However, detection of CIS and resection of all tumors are essential for risk stratification and clinical decision-making, and they prevent recurrence effectively [7]. To improve detection rates, WL imaging can be combined with additional imaging modalities. Photodynamic diagnosis (PDD) involves preoperative bladder instillation of a substrate of heme metabolism. Subsequently, protoporphyrin IX (PpIX) accumulates in malignant cells and emits red fluorescence under excitation with blue light [8]. PDD has shown a clear advantage over WL in detecting flat lesions. Up to 40% of all CIS can only be identified under PDD according to a recent meta-analysis [6,9]. Narrow-band imaging (NBI) uses light of defined wavelengths that are strongly absorbed by the hemoglobin in vessels of the bladder mucosa. This leads to enhanced vascular contrast (EVC) that helps to identify potentially malignant lesions with increased or abnormal vasculature. As for PDD, studies have shown that NBI has clear value in increasing the rate of BC detection [10]. We recently showed that tissue autofluorescence (AF) can also assist in distinguishing normal and malignant bladder-wall tissue in wide-field endoscopy [11]. Further imaging modalities such as digital contrast enhancement and near-infrared imaging are currently under evaluation [12]. All the various adjunct imaging modalities have individual drawbacks and limitations. Modalities that require specific illumination (PDD, NBI, and AF), for instance, can only be visualized separately, but not in parallel with or overlaid with WL. Hence, repetitive switching during TUR-B is required. Moreover, existing endoscopic systems cannot combine multiple imaging modalities [13]. Real-time multispectral imaging (rMSI) allows separate visualization
of multiple spectral components and thus can extract information not visible in WL images [14]. The aim of our study was to adapt an rMSI device for urethrocystoscopy to allow visualization and combination of multiple adjunct imaging modalities (PDD, EVC, AF) with WL imaging. This methodology may allow multiparametric cystoscopy (MPC) for detection of BC for the first time. 2.
Patients and methods
2.1.
Imaging setup
The general setup for rMSI consists of a camera unit, a light source, and a computer with a microcontroller board for control of the camera and light source (Fig. 1). A color scientific complementary metal-oxide-semiconductor camera (PCO AG, Kelheim, Germany) is used for imaging. The endoscope is mounted on the camera using a C-mount adapter (R. Wolf, Knittlingen, Germany). An optical multibandpass filter (Chroma Technology, Bellows Falls, VT, USA) is mounted on the C-mount adapter using a custom 3Dprinted mount and two-phase dental glue (Picodent Dental Produktions und Vertriebs, Wipperfürth, Germany). A modular LED light source (Omicron-laserage Laserprodukte, Rodgau, Germany) with peak wavelengths distributed over the visible range is used. The LEDs are individually filtered to match the bandpass properties of the filter in front of the camera sensor. The camera and light source are controlled via a microcontroller board interfaced with bespoke software running on a high-end personal computer, which is also used for basic image processing and displaying the video stream on the screen in real time during cystoscopy. For cystoscopy, sterilized PDD endoscopes (30 optics, Karl Storz, Tuttlingen, Germany) and sterilized liquid light guides (Karl Storz) are used. MSI is achieved via temporal multiplexing of WL, EVC, and PDD illumination, providing a video stream with a frame rate of approximately 20 Hz. 2.2.
Image processing
Frames of color imaging (WL) and total fluorescence (PDD) are white balanced, and the EVC frames are mapped to different RGB channels of the output image. In postprocessing (but not the live view), the contrast of WL and EVC frames is also increased by applying contrast-limited adaptive histogram equalization, and the saturation of the ECV and PDD images is also increased. The MP image is obtained via digital fusion of the EVC and PDD images (Fig. 2A). This is possible in real time, allowing for real-time MP imaging. For standardization, the MP images reviewed by urologists (see Section 2.4) were
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 E U RO P E A N U RO L O GY X X X ( 2 019 ) X X X – X X X
3
Fig. 1 – Schematic diagram of the real-time multispectral imaging system. The system consists of a scientific complementary metal-oxidesemiconductor camera with a C-mount, a video adapter with a glued optical filter, and a light source. The camera and light source are controlled via a microcontroller. A personal computer (PC) records the data and displays the videos to the surgeon.
generated after the intervention. The PpIX fluorescence (PpIX-F) and AF modalities are derived from the PDD modality via linear unmixing with an empirically determined unmixing matrix (Fig. 2B). Thus, the resulting component images are monochrome and are colored afterwards for better visualization.
that WL, EVC, PDD (and thus PpIX and AF, but not unmixed), and the merged MP image were available during cystoscopy. Images of all rMSI modalities analyzed were generated after interventions from the raw data using a standardized procedure as described above. 2.4.
2.3.
Image analysis and statistical evaluation
Patients and surgical procedure
Overall, ten patients scheduled for TUR of a cystoscopically diagnosed bladder tumor were enrolled. All patients provided written informed consent for participation in this pilot study, which was approved by the local ethics committee (approval number 2015-616N-MA). All patients received a bladder instillation of 85 mg of hexaminolevulinate (Hexvix, Ipsen Pharma, Boulogne-Billancourt, France) 1 h before the procedure. Surgery was performed under general or spinal anesthesia. MSI was used to scan the entire bladder for suspicious lesions. Video data were recorded synchronously. After MSI, conventional endoscopy using the D-Light System with a Tricam SL II camera and a Ch26 continuous-flow shaft with 30 lenses (Karl Storz) was performed. All suspicious lesions identified with the standard instrument were biopsied or resected for a standard histological examination. For bipolar resection, an Autocon II 400 SCB system (Karl Storz) was used. Initially, WL, EVC, and PDD were available in real time during MSI for all patients. During the course of the study we were able to further integrate real-time MP imaging, so
After surgery, the video sequences were reviewed by the operating surgeon to identify frames showing the resected lesions. Images of those lesions in all six modalities (WL, EVC, PDD, PpIX-F, AF, and MP) were independently reviewed in random order by two urologists and scored on a fourpoint Likert scale, with 0 denoting “not suspicious” and 3, “clearly suspicious”. The scores for the 27 malignant lesions (n = 324 scores) were further analyzed using STATA (StataCorp LLC, College Station, TX, USA). To tease out the effect of the modality under which a lesion was observed on the degree of suspicion it aroused, a multilevel mixed-effects ordered logistic regression (meologit command) was built with fixed effects for modality, images nested in lesions and controls for a patient, reviewer (surgeon), and whether the image contained a flat lesion. From the resultant model, the predicted probabilities of each score for each modality were estimated with 95% confidence intervals and compared pairwise for each score. The choice of the statistical model was motivated by the ordinal character of the dependent variable (score), multiple
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 4
E U RO P E A N U R O L O GY X X X ( 2 019 ) X X X – X X X
Fig. 2 – (A) Merger of enhanced vascular contrast (EVC) luminescence-based contrast and photodynamic diagnosis (PDD) fluorescence-based contrast into one multiparametric (MP) image. (B) Unmixing of full PDD fluorescence image into protoporphyrin IX fluorescence (PpIX-F) and autofluorescence (AF) components.
Table 1 – Patient and tumor characteristics for the pilot study population. Number
Sex
Age (yr)
Recurrence
Tumor size (cm)
Lesions (n)
Histology
1 2 3 4 5 6 7 8 9 10
M M M M M M F M M M
70 74 61 79 73 73 64 55 81 74
No Yes Yes No No Yes No Yes Yes No
<3 <3 <3 >3 >3 >3 >3 >3 >3 <3
2 3 4 1 2 10 1 1 4 2
Benign UCC UCC UCC UCC UCC SCC UCC UCC UCC
TMN
Grade
– pTa pTa pTa pT1 pTa >pT2 pTa pTa + Tis pTa
– LG LG LG HG LG HG HG LG + HG LG
M = male; F = female; UCC = urothelial cell carcinoma; SCC squamous cell carcinoma; HG = high grade; LG = low grade.
observations of the same lesion in different modalities, and possible patient- and reviewer-specific influences. The likelihood ratio test against the simpler ordinary logistic regression model was significant (p = 0.0055), indicating that there is sufficient variance between lesions to warrant a multilevel model. Robustness checks against the simpler model with clustered standard errors or fixed effects for lesion did not produce substantially different results. In addition, keeping in mind that all images contained a malignant lesion, we transformed the scores into a dichotomous variable, defining scores of 0 and 1 as failure, and scores of 2 and 3 as success for each modality in arousing surgeon suspicion. We coded failure as 1 and success as 0. A multilevel, mixed-effects logistic regression model was constructed with fixed effects for modality, images nested in lesions and controls for the reviewer, and flat lesion. This allowed us to estimate the predicted failure
rate for each modality with 95% confidence intervals and to obtain corresponding p values. The MP modality was dropped from the model because no case of failure was present in the data.
3.
Results
rMSI was successfully performed in ten patients (nine male, one female) with diagnosed bladder tumors referred for TUR-BT. Patient characteristics are displayed in Table 1. Half of the patients had a history of previous BC. Malignant pathology was found in nine patients. The median number of lesions per patient was two (range 1–10). The median interventional time was 31 (11–120) min and no intraoperative or postoperative complications occurred. Overall, 31 lesions were assessed via rMSI. The lesion characteristics
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 5
E U RO P E A N U RO L O GY X X X ( 2 019 ) X X X – X X X
Table 2 – Macroscopic and pathological description of the 30 lesions identified and semiquantitative rating results from two urologistsa. Patient
Lesion
Histology
TMN
1
1 2 1 2 3 1 2 3 4 1 1 2 1 2 3 4 5 6 7 8 9 10 1 1 1 2 3 4 1 2 3
BI BI UCC UCC UCC UCC UCC UCC BI UCC UCC UCC UCC UCC UCC UCC UCC UCC UCC UCC UCC UCC sUCC UCC UCC UCC BI UCC UCC UCC UCC
– – pTa pTa pTa pTa pTa pTa
LG LG LG LG LG LG
pTa pT1 pTa pTa pTa pTa pTa pTa pTa pTa pTa pTa pTa >pT2 pTa Tis pTa
LG HG HG LG LG LG LG LG LG LG LG LG LG HG HG HG LG
Tis pTa pTa pTa
HG LG LG LG
2
3
4 5 6
7 8 9
10
Grade
Location
4 4 5 5 5 7 5 5 4 1, 7 7 7 4 7, 2 7 7 8 6, 8 1 2, 4 4 8 8 8, 2 1 1 7 9 4 2 2
Architecture
WL
Flat sessile Flat Papillary Papillary Papillary Sessile Sessile Microsessile Microsessile Flat papillary Exophytic sessile Exophytic sessile Papillary Flat papillary Papillary Small papillary Papillary Flat papillary Small papillary Flat Flat Small sessile Large sessile Flat papillary Flat papillary Flat papillary Flat Flat Papillary Papillary Papillary
U1 0 1 3 3 3 2 2 0 3 3 3 3 3 3 1 3 3 3 1 1 2 3 2 3 3 1 0 3 3 3
U2 1 2 3 2 2 3 3 0 0 3 2 3 3 3 2 1 3 3 3 1 0 2 2 1 2 3 2 1 2 3 2
EVC U1 1 2 3 3 3 2 2 0 1 3 3 3 3 3 3 2 3 3 3 1 2 3 3 2 2 3 1 1 3 3 2
U2 1 2 2 3 3 1 2 0 1 3 2 2 3 3 3 2 3 2 3 2 1 3 2 2 2 3 2 1 3 3 2
PDD U1 1 0 2 2 3 2 2 2 0 3 3 3 3 3 3 3 2 2 3 3 2 3 2 1 3 2 2 2 3 3 3
U2 1 1 3 2 3 2 1 2 1 3 3 3 3 3 3 3 2 2 3 3 2 3 1 1 3 2 2 3 2 2 2
PpIX-F U1 0 0 2 2 2 2 1 2 0 3 3 3 3 3 3 2 2 2 3 3 1 3 1 1 3 2 0 2 3 3 3
U2 0 0 3 2 2 2 2 2 0 3 3 3 3 3 3 2 3 2 3 3 1 3 1 1 3 2 0 1 2 3 3
AF U1 0 2 3 1 2 2 2 2 0 3 1 1 1 3 3 3 2 1 3 3 2 3 1 1 2 1 1 1 3 2 3
MP U2 0 1 3 1 1 2 1 2 0 3 2 2 3 3 2 3 2 1 3 2 2 2 2 2 1 1 2 2 2 2 2
U1 1 2 3 3 3 3 2 2 0 2 3 3 3 3 3 3 3 3 3 3 2 3 3 3 3 3 1 3 3 3 3
U2 1 2 3 3 3 3 3 2 1 2 3 3 3 3 3 3 3 3 3 3 2 3 3 2 3 3 2 3 3 3 3
BI = benign inflammation; UCC = urothelial cell carcinoma; sUCC = squamous cell urothelial carcinoma; HG = high grade; LG = low grade; WL = white light; EVC = enhanced vascular contrast; PDD = photodynamic diagnosis; PpIX-F = protoporphyrin IX fluorescence; AF = autofluorescence; MP = multiparametric. a The two physicians (U1 and U2) scored lesion suspicion on the different imaging modalities on a scale from 0 to 3.
are summarized in Table 2. Histopathology revealed a malignant tumor in 27 lesions, of which the majority were pTa low grade (n = 22). A high-grade tumor was identified in five lesions, of which two were CIS and one was a muscleinvasive, squamous-like tumor. rMSI visualized the bladder mucosa using six different modalities: WL, EVC, PDD, PpIX-F, AF, and MP. No technical difficulties occurred. All lesions identified with the rMSI system were judged suspicious on standard PDD, and therefore biopsies were taken. Furthermore, standard PDD did not identify any additional lesions previously not detected with rMSI. Fig. 3 shows different lesions in all modalities. Lesions 1 and 2 are typical papillary bladder tumors that are clearly visible in all modalities. Both tumors have a positive PpIX-F and a negative AF signal. Lesion 3 is a flat tumor at the bladder trigonum. Under WL, the lesion is hardly demarcated from the surrounding mucosa. EVC mode allows identification due to differences between the tumor and the adjacent mucosa. In PDD mode, the tumor is clearly highlighted. Under separate PpIX-F and AF displays, the putatively malignant tissue exhibits strong PpIX-F and weak AF, whereas normal mucosa exhibits strong AF signals. In the MP image, the specific characteristics of the single modalities are overlaid on the WL image. Lesion 4 comprises very small papillary lesions on the back wall of the bladder. The different modalities again contribute to
clear demarcation of the malignant lesions, and they are combined in the MP image. Lesion 5 is CIS on the left bladder wall. Positive PpIX-F and weak AF arouse suspicion of a malignant process. Images of all 27 malignant lesions in each modality were rated by two urologists on a Likert scale from 0 (not suspicious) to 3 (clearly suspicious). The results are reported in Table 2 and presented graphically in Fig. 4. MP was the only modality for which all images were scored as 2 or 3; other modalities also received lower ratings. In addition, the fraction of images scoring 3 was highest for the MP modality (>0.8). Multilevel ordinal logistic regression for these ratings revealed a significant positive effect of the MP modality on score (p < 0.001) and a negative effect of the AF modality (0.046). Images showing flat lesions in general received lower scores (p < 0.001). The predicted probability of completely missing a lesion (score 0 on the Likert scale) was significantly greater than zero for all modalities other than MP (p = 0.128; Fig. 4). In particular, the predicted probability of WL images being scored 0 was 0.0234 (p = 0.042), which was significantly higher (p = 0.045) than for MP (0.002; p = 0.128). The probability of an MP image being scored 1 was 0.016, which was significantly lower than all other modalities (all p < 0.001) and in particular PDD (0.100), EVC (0.115), and WL (0.138). MP images had the highest predicted probability of
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 6
E U RO P E A N U R O L O GY X X X ( 2 019 ) X X X – X X X
Fig. 3 – Examples of lesion images in all modalities. Lesions 1 and 2 are typical papillary tumors and are clearly visible in all modalities. Both tumors show strong protoporphyrin IX fluorescence (PpIX-F) and negative autofluorescence (AF). Final histology was pTa low grade. Lesion 3 is a small and flat tumor (pTa low grade) at the bladder trigonum and is barely detectable with white light (WL). Its abnormal vasculature is demarcated on enhanced vascular contrast (EVC) mode. It also shows positive PpIX-F and low AF. All features are overlaid in the multiparametric (MP) mode. Lesion 4 consists of very small papillary lesions (pTa low grade). Comparable to lesion 3, the single modalities contribute to clearer identification of the tumors. Lesion 5 is a flat lesion with the final pathology of carcinoma in situ. It has an abnormal vasculature highlighted by EVC with positive PpIX-F and negative AF. The MP mode clearly demarcates the lesions and their full extent.
receiving a score of 3 (clearly suspicious). The second highest was PDD at 0.567, with WL at 0.474. The probabilities of each modality failing to arouse substantial suspicion as predicted by the logistic regression model for the dichotomous variable are presented in Fig. 5. With no failures, MP was excluded from this analysis. Among the other modalities, all had a probability of failure of significantly greater than zero (p < 0.001 for WL and AF, others up to p = 0.036). Only PDD had a significantly lower probability of failure than WL (0.075 vs 0.200; p = 0.039). Failure was more likely for images containing flat lesions (p = 0.006). There was no statistically significant difference between the two reviewers (p = 0.852). 4.
Discussion
WL endoscopy currently remains the imaging gold standard in endourology. It offers high sensitivity and specificity for most papillary bladder tumors [4,5]. However, it has major limitations in the detection of flat and small lesions, up to 50% of which are missed under WL [6]. Additional imaging modalities such as PDD and NBI can add information and support urologists in identifying malignant urothelial lesions [13]. Various studies have shown a benefit of PDD and NBI in increasing detection
rates. In a meta-analysis, Burger et al [9] found that 40% of flat lesions can only be detected when using PDD. Eventually, use of PDD during resection might even contribute to an increase in recurrence-free survival [6]. Similar data exist for NBI [10]. Using AF emitted by endogenous fluorophores in the mucosal layer, epithelial cancer could be detected in different endoscopic fields (eg, bronchoscopy and colonoscopy) [15,16]. We recently demonstrated that AF may also improve the rate of BC detection [11]. Unfortunately, all these additional modalities have individual drawbacks and they can only be applied separately. So far, a combination or even simultaneous application has not been possible. We present the first human application of MP endoscopy in urology. Using an rMSI camera and software system, we could visualize the bladder mucosa in up to five different modalities simultaneously. Fusion of specific information from these modalities in one endoscopic image was feasible. This technical approach allowed MPC with astounding and diverse potential. Beside demonstrating feasibility, our preliminary data suggest a better detection rate, especially for small and flat malignant lesions of the bladder wall. In our pilot study, all malignant lesions were rated as suspicious using the MP modality, whereas observers rated some malignant lesions as not or only a little suspicious
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 E U RO P E A N U RO L O GY X X X ( 2 019 ) X X X – X X X
7
Fig. 4 – Fraction and predicted probability with 95% confidence interval of each score for each modality. Fractions were calculated directly from histograms of the scores assigned by reviewers to images containing malignant lesions. Predicted probabilities are from the multilevel ordinal logistic regression model after estimation. MP = multiparametric; WL = white light; EVC = enhanced vascular contrast; PDD = photodynamic diagnosis; PpIX = protoporphyrin IX fluorescence; AF = autofluorescence.
when using the single modalities. We thus combined the individual benefits of each modality in a merged image that compensated for the limitations of the individual modalities. Some malignant lesions may not appear suspicious in some modalities. Data from a recent study suggest that more malignant flat lesions might be detected when PDD and NBI are used subsequently [17]. PDD has a failure rate in routine clinical practice, and some malignant tumors show no or little fluorescence. In an MP approach, this could be compensated by other modalities such as EVC and AF. Moreover, PDD/PpIX-F, EVC, and AF are excellent complements to each other: PDD/PpIX-F gives information on the metabolic activity of the tissue by highlighting mucosa with high uptake of 5-aminolevulinic acid [18]. EVC draws the observer’s eye to highly vascularized and thus potentially malignant tissue [10]. AF provides information on thickening and abnormal composition of the epithelial layer. We previously found that AF has a high negative predicitve value, suggesting that a strong AF signal reliably indicates a normal healthy mucosa [11]. Consequently, AF is a suitable supplement to PDD/PpIX-F and EVC, which predominantly highlight potentially malignant tissue. This underlines the unique advantages of the MP approach. So far, some providers have introduced endoscopic platforms that
present different images to the urologist, such as the SPIES system (Karl Storz), which uses digital image processing and contrast enhancement to highlight different aspects of the image (eg, vasculature, depth, and illumination) [12]. However, this approach is limited to data acquired only from WL images, and no additional information on metabolism or epithelial composition is displayed for the physician. By contrast, the objective of MP imaging is to combine modalities with different metabolic and anatomic implications in a true MP approach. MP imaging technology enhances WL cystoscopy and offers high diagnostic potential not only for better BC detection but also for more complete TUR. Most surgeons currently use additional imaging modalities to detect putatively malignant lesions, but they prefer to resect under WL. This carries the risk of imprecise tumor resection, especially at the edges of flat and diffusely growing lesions. Incomplete resection is believed to be at least partly responsible for a tumor detection rate of up to 40% during secondary TUR-BT [19]. By providing WL images and additional imaging information (eg, fluorescence) during the resection in overlay and in parallel, the MP cystoscopy technology might improve resection and reduce the remaining cancerous tissue. The proof of the assumption
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 8
E U RO P E A N U R O L O GY X X X ( 2 019 ) X X X – X X X
Fig. 5 – Predicted probability of failure for each modality with 95% confidence interval obtained from multilevel mixed-effects logistic regression on the dichotomous variable after estimation. MP = multiparametric; WL = white light; EVC = enhanced vascular contrast; PDD = photodynamic diagnosis; PpIX = protoporphyrin IX fluorescence; AF = autofluorescence.
requires further studies. However, there might be a predisposition to a higher rate of false-positive findings, especially during MP TUR-BT, since abnormal findings in any modality might induce a decision to biopsy. In the future, training and experience in MP cystoscopy as well as computer-aided systems may support surgeons in avoiding unnecessary biopsies. Future MPC perspectives could include the application of fluorescence-labeled targeted antibodies to specifically label BC. A recent ex vivo study used CD47 antibodies to mark BC for blue-light endoscopy and confocal endomicroscopy in ex vivo radical cystectomy specimens [20]. Higher sensitivity for detection of malignant lesions and molecular marking may also help to identify variants of BC that are diagnostically challenging. Moreover, specificity might increase, which would improve diagnostic accuracy for CIS or recurrent BC following prior intravesical immunotherapy with bacillus Calmette-Guérin [20]. MP endoscopy in theory allows combination with antibody-based detection and could thus potentially further improve detection rates. Furthermore, the inclusion of other potential molecular markers such as EGFR and PSCA—with the use of differentially labeled antibodies—could be useful for specifying findings, in line with the idea of molecular endoscopy [21,22]. In the current MP imaging implementation, differentiation (and thus combination of up to three different fluorophores) is feasible, which opens the door to a diverse molecular approach. Molecular endoscopy needs a multistep preclinical evaluation process, for which an in vivo model has been established by our group [23]. Our study presents initial clinical data on a new imaging device. It is limited by its proof-of-concept design, with a relatively small number of patients and lesions assessed. Although we have demonstrated the feasibility of MP endoscopy for detection of BC, the added value of MPC in clinical routine needs to be validated. Moreover, resection was performed using conventional endoscopic equipment (WL and PDD) following MPC, and thus the value of MPC during tumor resection needs to be examined further. Ultimately, MPC needs to be compared to existing state-ofthe-art endoscopic imaging modalities, preferably in a head-to-head study.
5.
Conclusions
In this pilot study, we enhanced cystoscopy by implementing rMSI in endourology and by combining different imaging modalities. This allowed us to perform MPC for the first time for the diagnosis of malignant bladder lesions. Up to five different modalities (WL, EVC, PDD, PpIX-F, and AF) were displayed simultaneously and combined into an MP image. Our preliminary results suggest better detection of bladder tumors using MPC. This technology may further complement fluorescent signals, for example, by using fluorescence-labeled specific antibodies for molecular endoscopy. Author contributions: Maximilian C. Kriegmair had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Kriegmair, Bolenz, Deliolanis. Acquisition of data: Kriegmair, Rother, Theuring. Analysis and interpretation of data: Kriegmair, Grychtol, Bolenz. Drafting of the manuscript: Kriegmair, Rother, Grychtol. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: Grychtol. Obtaining funding: Bolenz, Deliolanis. Administrative, technical, or material support: Rother, Theuring, Grychtol, Deliolanis, Günes. Supervision: Bolenz, Deliolanis. Other (video): Rother, Kriegmair. Financial disclosures: Maximilian C. Kriegmair certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Nikolaos C. Deliolanis is a co-inventor for multispectral imaging patent applications. The remaining authors have nothing to disclose.
Funding/Support and role of the sponsor: This work was supported by funding from the German Ministry for Education and Research (GO-Bio Project 031B0219). The sponsor played no direct role in the study.
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024
EURURO-8546; No. of Pages 9 E U RO P E A N U RO L O GY X X X ( 2 019 ) X X X – X X X
Appendix A. Supplementary data
9
[11] Kriegmair MC, Honeck P, Theuring M, Bolenz C, Ritter M. Wide-field autofluorescence-guided TUR-B for the detection of bladder cancer:
The Surgery in Motion video accompanying this article can be found in the online version at doi:https://doi.org/10. 1016/j.eururo.2019.08.024.
a pilot study. World J Urol 2018;36:745–51. [12] Kamphuis GM, de Bruin DM, Brandt MJ, et al. Comparing image perception of bladder tumors in four different Storz Professional Image Enhancement System modalities using the iSPIES app. J Endourol 2016;30:602–8.
References
[13] Lerner SP, Goh A. Novel endoscopic diagnosis for bladder cancer. Cancer 2015;121:169–78.
[1] Antoni S, Ferlay J, Soerjomataram I, Znaor A, Jemal A, Bray F. Bladder
[14] Dimitriadis N, Grychtol B, Theuring M, Behr T, Sippel C, Deliolanis
cancer incidence and mortality: a global overview and recent
NC. Spectral and temporal multiplexing for multispectral fluores-
trends. Eur Urol 2017;71:96–108. [2] Wong MCS, Fung FDH, Leung C, Cheung WWL, Goggins WB, Ng CF. The global epidemiology of bladder cancer: a joinpoint regression
cence and reflectance imaging using two color sensors. Opt Express 2017;25:12812–29. [15] He Q, Wang Q, Wu Q, Feng J, Cao J, Chen BY. Value of autofluores-
analysis of its incidence and mortality trends and projection. Sci Rep
cence imaging videobronchoscopy in detecting lung cancers and
2018;8:1129.
precancerous lesions: a review. Respir Care 2013;58:2150–9.
[3] Burger M, Catto JW, Dalbagni G, et al. Epidemiology and risk factors of urothelial bladder cancer. Eur Urol 2013;63:234–41.
[16] Li BH, Xie SS. Autofluorescence excitation-emission matrices for diagnosis of colonic cancer. World J Gastroenterol 2005;11:3931–4.
[4] Borin JF, Abdelshehid CS, Clayman RV. Comparison of resolution,
[17] Drejer D, Beji S, Oezeke R, et al. Comparison of white light, photo-
contrast, and color differentiation among fiberoptic and digital
dynamic diagnosis, and narrow-band imaging in detection of car-
flexible cystoscopes. J Endourol 2006;20:54–8. [5] Cina SJ, Epstein JI, Endrizzi JM, Harmon WJ, Seay TM, Schoenberg MP. Correlation of cystoscopic impression with histologic diagnosis of biopsy specimens of the bladder. Hum Pathol 2001;32:630–7. [6] Rink M, Babjuk M, Catto JW, et al. Hexyl aminolevulinate-guided fluorescence cystoscopy in the diagnosis and follow-up of patients with non–muscle-invasive bladder cancer: a critical review of the current literature. Eur Urol 2013;64:624–38. [7] Soukup V., Capoun O., Cohen D., et al. Risk stratification tools and
cinoma in situ or flat dysplasia at transurethral resection of the bladder: the DaBlaCa-8 study. Urology 2017;102:138–42. [18] Nakai Y, Tatsumi Y, Miyake M, et al. Expression of ferrochelatase has a strong correlation in protoporphyrin IX accumulation with photodynamic detection of bladder cancer. Photodiagn Photodyn Ther 2016;13:225–32. [19] Schwaibold HE, Sivalingam S, May F, Hartung R. The value of a second transurethral resection for T1 bladder cancer. BJU Int 2006;97:1199–201.
prognostic models in non-muscle-invasive bladder cancer: a critical
[20] Pan Y, Volkmer JP, Mach KE, et al. Endoscopic molecular imaging of
assessment from the European Association of Urology Non–Muscle-
human bladder cancer using a CD47 antibody. Sci Transl Med
Invasive Bladder Cancer Guidelines Panel. Eur Urol Focus. In press. https://doi.org/10.1016/j.euf.2018.11.005.
2014;6:260ra148. [21] Rebouissou S, Bernard-Pierrot I, de Reynies A, et al. EGFR as a
[8] Kriegmair M, Baumgartner R, Knuchel R, Stepp H, Hofstadter F,
potential therapeutic target for a subset of muscle-invasive bladder
Hofstetter A. Detection of early bladder cancer by 5-aminolevulinic
cancers presenting a basal-like phenotype. Sci Transl Med
acid induced porphyrin fluorescence. J Urol 1996;155:105–9.
2014;6:244ra91.
[9] Burger M, Grossman HB, Droller M, et al. Photodynamic diagnosis of
[22] Yuan R, Rao T, Cheng F, et al. Quantum dot-based fluorescent probes
non-muscle-invasive bladder cancer with hexaminolevulinate cys-
for targeted imaging of the EJ human bladder urothelial cancer cell
toscopy: a meta-analysis of detection and recurrence based on raw data. Eur Urol 2013;64:846–54.
line. Exp Ther Med 2018;16:4779–83. [23] Günes C, Meessen S, Rother J, et al. Development of a rodent model
[10] Zheng C, Lv Y, Zhong Q, Wang R, Jiang Q. Narrow band imaging
for preclinical evaluation of multiple contrast agents and real-time
diagnosis of bladder cancer: systematic review and meta-analysis.
multispectral imaging in bladder cancer. Eur Urol Suppl 2019;18:
BJU Int 2012;110:E680–7.
e1955.
Please cite this article in press as: Kriegmair MC, et al. Multiparametric Cystoscopy for Detection of Bladder Cancer Using Realtime Multispectral Imaging. Eur Urol (2019), https://doi.org/10.1016/j.eururo.2019.08.024