Presurgical Planning for Supratentorial Lesions with Free Slicer Software and Sina App

Accepted Manuscript Presurgical planning for supratentorial lesions with free Slicer software and Sina app Ji-gang Chen, Kai-wei Han, Dan-feng Zhang, Zhen-xing Li, Yi-ming Li, Li-jun Hou PII:

S1878-8750(17)31043-4

DOI:

10.1016/j.wneu.2017.06.146

Reference:

WNEU 6017

To appear in:

World Neurosurgery

Received Date: 28 February 2017 Revised Date:

21 June 2017

Accepted Date: 24 June 2017

Please cite this article as: Chen J-g, Han K-w, Zhang D-f, Li Z-x, Li Y-m, Hou L-j, Presurgical planning for supratentorial lesions with free Slicer software and Sina app, World Neurosurgery (2017), doi: 10.1016/j.wneu.2017.06.146. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Title: Presurgical planning for supratentorial lesions with free Slicer software and Sina app Authors: Ji-gang Chen1, Kai-wei Han1, Dan-feng Zhang1, Zhen-xing Li, Yi-ming Li,

RI PT

Li-jun Hou (1These authors contributed equally to this work and should be considered co-first

Affiliation:

Department

of

Neurosurgery,

Shanghai

neurosurgical

institute,

M AN U

Changzheng Hospital, Shanghai, China.

SC

authors)

Corresponding Author’s name and complete mailing address:

TE D

Li-jun Hou, Department of Neurosurgery, Shanghai neurosurgical institute, Changzheng Hospital, Shanghai, China. Phone: 0086-021-81885671

EP

Fax: 0086-021-81885698

AC C

Email: [email protected]

Yi-ming Li, Department of Neurosurgery, Shanghai neurosurgical institute, Changzheng Hospital, Shanghai, China. Phone: 0086-021-81885671 Fax: 0086-021-81885698 Email: [email protected]

ACCEPTED MANUSCRIPT Background: Neuronavigation system is widely used in the localization of intracranial lesions with satisfactory accuracy. However, it’s expensive and difficult to learn. Therefore, simple and practical augmented reality (AR) system using mobile

RI PT

devices might be an alternative technique. Objective: We aim to introduce a mobile AR system for the localization of supratentorial lesions. Its practicability and accuracy were examined by clinical

SC

application in patients and comparison with standard neuronavigation system.

M AN U

Methods: Three-dimensional (3D) model including lesions is created using 3D Slicer. 2D image of this 3D model was obtained and overlapped on patients’ head using Sina app. Registration was conducted with the assistance of anatomical landmarks and fiducial markers. Center of lesion projected on scalp was identified with our mobile

TE D

AR system and standard neuronavigation system, respectively. Distance difference between centers identified by these two systems was measured. Result: Our mobile AR system was simple and accurate in the localization of

EP

supratentorial lesions with a mean distance difference of 4.4 ± 1.1 mm. Registration

AC C

added on an average of 141.7 ± 39 seconds to operation time. There was no statistically significant difference for the required time among three registrations (P=0.646).

Conclusion: The mobile AR system presents an alternative technology for image-guided neurosurgery and proves to be practical and reliable. The technique contributes to optimal presurgical planning for supratentorial lesions, especially in the absence of neuronavigation system.

ACCEPTED MANUSCRIPT Key words: Augmented reality; Mobile application; Neuronavigation; 3D Slicer Introduction In neurosurgical operations, precise presurgical planning of surgical approach

RI PT

and craniotomy is of great importance, which depends much on detailed preoperative localization of intracranial lesions. Experienced neurosurgeons learn to correlate patient’s anatomy with presurgical two-dimensional (2D) images and localize the

SC

intracranial lesions. However, the accuracy of such methods is sometimes not

M AN U

satisfactory1 since it is subjective and closely related to clinical experience. Neuronavigation system for intraoperative image guidance improves surgical accuracy and surgeons’ orientation, reducing iatrogenic injury to the patients.2-4 As an established technology, neuronavigation has been widely used in neurosurgery.1, 2, 5, 6

TE D

However, the neuronavigational system is expensive and relativlely difficult to learn because it requires hand-eye coordination of the neurosurgeon to operate according to images in the monitor.2 Therefore, it is necessary to develop simple and practical

EP

techniques for the fusion of virtual images and real surgical field.

AC C

Augmented reality (AR) is a fast-developing technology which combines the reality and virtual images, allowing the presentation of virtual images in real environment.2, 7 Aside from implementation of AR using expensive image guidance systems and surgical microscopes, some authors have described alternative low-cost solutions using mobile devices in neurosurgical settings, which are referred as mobile AR.8, 9 This technology renders AR as much more convenient, affordable and popular. In current study, we report a simple and practical mobile AR system for

ACCEPTED MANUSCRIPT presurgical planning of scalp flap and localization of supratentorial lesions, which works mainly with a free Slicer software and a smartphone app. Methods

RI PT

Patients From October 2015 to October 2016, a group of 16 patients with supratentorial lesions were prospectively admitted into our department. This study was approved by

SC

Ethics Committee of the Changzheng Hospital and signed informed consent for each

M AN U

patients was available.

Three-dimensional (3D) modeling and segmentation with free Slicer software All the patients, independent of their sex, had their hair fully shaved. Before

TE D

magnetic resonance imaging (MRI) scan, 5 vitamin E soft capsules (100mg, Xinchang Pharmaceutical Factory, Zhejiang, China) were pasted around the lesions with self-adhesive films according to preoperative head computed tomography (CT)

EP

presentations. All MRI scan were performed with 1.5 Tesla magnetic resonance

AC C

scanner (Signa Excite; GE Medical Systems, Milwaukee, WI, USA), and data of plain and enhanced MRI TI phases were exported as Digital Imaging and Communications in Medicine (DICOM) medical images. We used sequence parameters as follows: repetition time, 9.6ms; echo time, 4.3ms; flip angle, 15 degrees; matrix, 256*256; slice thickness, 0.8mm; field of view, 160*160mm; sequence, gadolinium-enhanced three-dimentional fast spoiled gradient-echo (3D-FSPGR). Segmentation and 3D modeling were performed using 3D Slicer (3D Slicer 4.0;

ACCEPTED MANUSCRIPT Surgical Planning Laboratory, Harvard University, USA) on the basis of DICOM images of included patients. Models of the superior sagittal sinus with cortical veins, markers of vitamin E soft capsules, and lesion were segmented out and built,

RI PT

respectively. A small sphere model with a diameter of 2mm was constructed and marked with blue. It was located at the center of scalp projection area of lesion. This sphere model could be localized by neuronavigation system later and used as a

SC

reference point to check the accuracy of our AR system. All the models were exported

M AN U

in stereolithography (STL ) format. After the 3D model was rotated to a proper view angel (superior view for the frontal lesions and lateral view for the parietal lesions), 2D image of this view angle was then saved and exported. Two neuroradiologists (Chen JG and Han KW) reviewed the MRI data and conducted reconstruction

TE D

independently.

Import image into Sina app

EP

Sina is an Android app available in Google Play Store. The primary aim is to

AC C

assist in intraoperative neurosurgical planning. After 2D image of the reconstructed model was imported into the smartphone (HUAWEI honor 6 plus, Android 4.4.2), we launched Sina app and loaded the 2D image from image gallery. Then this image showed up on the phone screen as hemi-transparent and overlapped on the live feed from the camera. Since images in Sina couldn’t be modified, we superimposed the image on patients’ head manually by adjusting the distance between the device and patient’s head.

ACCEPTED MANUSCRIPT

Registration Patients laid down in a supine position after anesthesia. The midline of head was

RI PT

used as an anatomical landmark and signed with a marker pen preoperatively. 5 vitamin E capsules on the head were taken as fiducial markers. Sina was launched, and 2D image of the reconstructed model was loaded as the first step. Using the

SC

fiducial markers and midline as guides, the 2D image was overlapped on the head.

M AN U

Operator held the smartphone manually and adjusted the distance between device and head to make sure that the fiduical markers and midline could match with the corresponding markers and superior sagittal sinus on the image. In order to reduce error in the procedure, registration process was repeated three times for each patient

TE D

by three authors (Zhang DF, Li YM, Hou LJ) independently. Registration time was recorded and defined as time needed from starting the Sina app to the 2D image matching with patient’s head. After each registration, the center of lesion projected on

EP

scalp was marked as a dot directly by an assistant for the first two times and the

AC C

contour as well as center of lesion was drawn on patients’ head simultaneously for the third time.

Comparing the accuracy of Sina with standard neuronavigation system Standard neuronavigation was conducted for every patient using neuronavigation system (Brainlab Kick 1.0, Brainlab AG, Feldkirchen, Germany). The DICOM images and all the models constructed in Slicer previously were imported into

ACCEPTED MANUSCRIPT neuronavigation software. Standard registration of the neuronavigation system was performed according to the system’s protocol. The blue small sphere model was firstly located with the precalibrated pointer in the neuronavigation system and

RI PT

acquired as a reference point, which was automatically set as the “view center” in the neuronavigation system. Then the “Freeze” function of neuronavigation system was used, which enabled the measurement of distance between the reference point and the

SC

tip of the precalibrated pointer. Here we used this method to measure distance

M AN U

between the reference point localized by neuronavigation system and the point marked by AR system. The accuracy of Brainlab system in measuring distance was 0.1mm. The distance difference was then averaged from 3 available values.

TE D

Statistical analysis

Continuous data were expressed as the mean and standard deviation (SD). Friedman test was used to compare the distance differences and required times of

EP

three registrations. Statistical analyses were conducted using SPSS version 22.0 (IBM,

AC C

Inc., Chicago, IL, USA ). The significance level was defined as 0.05. Result

The mean age of our patients was 57.4 ± 8.7 years old (range, 43 to 74) and 8 of

them were male. 93.75% (15/16) of the patients were diagnosed with meningioma, with only 1 patient being glioma. 7 lesions were in the frontal lobe, 6 in the parietal lobe, 2 in both of the frontal and parietal lobe and 1 in the temporal lobe. (Table) The mean total registration time was 141.7 ± 39.4 seconds with the range of

ACCEPTED MANUSCRIPT 89-224 seconds. The mean time for the first registratoin was 48.6 ± 15.0 seconds (range: 30-79), for the second registration was 45.8 ± 12.9 seconds (range: 30-75) and for the third was 47.3 ± 12.9 seconds (range: 30-70). There was no statistically

RI PT

significant difference for the three registration times (P=0.646). Compared with standard neuronavigation system, our mobile AR system was accurate with a mean ± SD of 4.4± 1.1 mm and range of 2.3-6.3 mm.(Table) No statistically significant result

SC

was detected in distance difference of three registrations (P=0.153).

M AN U

Discussion

The beginning of AR can be traced back as early as 1960s.10, 11 As a great interest in health care, AR has been used in surgery for several decades, especially in neurosurgery where a combination of CT data and operating microscope was used in

TE D

stereotaxic operation to allow accurate and safe neuronavigation.12 Nowdays, AR is increasingly attractive as a popular technology due to the proliferation of smartphones.1, 8, 9 It’s being widely used with imaging technologies to superimpose

EP

MRI or CT data on patients during operation, offering surgeons navigational

AC C

information.13 Various AR methods for scalp localization have been reported using the anatomic structures, stereotactic frames, or frameless surgical navigators.2, 5 However, scalp localization should be accurate and simple. In this study, we presented a reliable and simple AR technique, which would be useful for presurgical planning and image-guided neurosurgery. Our method was a typical example of mobile AR implementation1,

8, 9

and

potentially carried the discussed advantages of these techniques by using Sina app

ACCEPTED MANUSCRIPT combined with Slicer software to assist in locating supratentorial lesions. In the present study, we used fiducial markers to increase the navigation accuracy. It was commonly applied in clinical practice and had been reported to err by 1.5mm to 4mm

RI PT

by various authors.14, 15, 16 On the basis of data from this study, the mean distance difference was 4.4 ± 1.1 mm, with a maximum deviation of 6.5 mm. Previous Sina app study which used anatomic structures such as the coronal suture and midline as

SC

the natural markers showed a deviation of 10.2 ± 2 mm.1 However, artificial markers

M AN U

were not used in this study. Our findings confirmed that AR would have a higher accuracy level if used with other registration models, such as adhesive or bone-implanted fiducials.17 This conclusion was consistent with several other studies.14, 18, 19, 20 Another important factor contributing to the accuracy of our method

TE D

was the use of 3D Slicer software for reconstruction. The reconstructed 3D image could be rotated to any view angle and we could get a precise 2D image of proper view angle from the 3D reconstructed model without any tilt. According to a similar

EP

mobile AR study using Sina app without 3D reconstruction, 1 degree of tilt could

AC C

cause about 2 mm of targeting deviation.1 What’s more, the angle of MRI scanning was determined manually by radiographer and axial MRI images might not be exactly vertical to the floor. In this way, the deviation would increase evidently if the head position was not in the right way during the MRI scanning. This deviation could be completely eliminated by 3D reconstruction of MRI image data. In our study, we reconstructed the cortical veins together with supratentorial lesions. Thus the relationship between two of them was clearly visualized after

ACCEPTED MANUSCRIPT registration. The basic craniotomy and skin flap planning could then be decided according to the localization of lesions and veins. Moreover, these cortical veins could be used as landmarks to seek lesions after dural opening. This would also be helpful

veins, such as the cerebral parafalx lesions.

RI PT

for venous protection in surgical process, especially for lesions closely related to

Theoretically, we can reconstruct a model which contains the sulci, gyri and

SC

cortical vessels with Slicer at the same time. After dural opening, we can coregister

M AN U

the model with surgical field of the cortex surface. It would likely be a useful neuronavigational aid during operations of small lesions, especially for those shallow subcortical ones. However, considering the brain shift after dural opening, the second registration would not be accurate enough. In this way, some authors proposed

TE D

methods such as intraoperative image updating to compensate for brain shift.21 Simplicity and availability of the device and software are the evident advantages as well. Compared with other AR solutions, our method achieved similar effects with

EP

simplified MRI image processing, registration procedure and AR image generation. It

AC C

does not need additional cost or technical complexity other than few markers, an android phone with a free App. Sina neurosurgical assist has been used in several studies for AR and proved to be effective with acceptable deviation.22, 23 Moreover, 3D Slicer is a free open source software popular for 3D modeling and has been applied for reconstruction in some AR studies.9, 24 Since our method is inexpensive and convenient with satisfactory accuracy, it could be used in the developing areas where neuronavigation system is inaccessible.

ACCEPTED MANUSCRIPT There are some limitations to this study which should be discussed. Though our method is reliable for supratentorial lesions, this does not apply exactly to infratentorial lesions. The success of our mobile AR system depends much on the

RI PT

definition obtained in the MRI scan reconstruction. The posterior fossa anatomy varies from supratentorial anatomy in a way that shallow sulci of cerebellar folia is difficult to identify in reconstructed model. It also has an almost parallel orientation,

SC

which makes registration unreliable. Moreover, cortical veins of the posterior fossa

M AN U

are of small diameter and difficult to reconstruct on 3D Slicer. All of these variables might explain why we are unable to perform our AR system in the infratentorial region.

Our method was only tested by a limited number of cases in a single center.

TE D

Multi-center studies with larger sample sizes are required to assess the feasibility and clinical application of this new method. Study on a large population by different users with various background might produce different results.

EP

In addition, manual image overlaying and lack of feedback about the precision of

AC C

overlap should be considered as the disadvantages of our method. Incomplete overlapping of the images with head may be the major source of deviation in this setting. The performers do not need much training in overlaying. However, the potential effect of user’s experience in using this app has not been addressed either.

Reference 1. Eftekhar BA. Smartphone App to Assist Scalp Localization of Superficial

ACCEPTED MANUSCRIPT Supratentorial Lesions--Technical Note. World neurosurgery. 2016;85:359-63. 2. Besharati Tabrizi L, Mahvash M. Augmented reality-guided neurosurgery: accuracy and intraoperative application of an image projection technique. Journal of

RI PT

neurosurgery. 2015;123(1):206-11. 3. Roessler K, Hofmann A, Sommer B, Grummich P, Coras R, Kasper BS, et al. Resective surgery for medically refractory epilepsy using intraoperative MRI and

SC

functional neuronavigation: the Erlangen experience of 415 patients. Neurosurgical

M AN U

focus. 2016;40(3):E15.

4. Zhang F, Hong W, Guo Y, Guo Q, Hu X. Multimodal Neuronavigation in Microsurgery Resection of BrainStem Tumors. The Journal of craniofacial surgery. 2016;27(8):769-772.

TE D

5. Mahvash M, Besharati Tabrizi L. A novel augmented reality system of image projection for image-guided neurosurgery. Acta neurochirurgica. 2013;155(5):943-7. 6. Zhang Y, Alaref R, Fu H, Yang Y, Feng Y, Zhao C, et al. Neuronavigation-assisted

EP

aspiration and electro-acupuncture for Hypertensive Putaminal Hemorrhage -a

AC C

suitable technique on hemiplegia rehabilitation. Turkish neurosurgery. 2016. 7. Tang SL, Kwoh CK, Teo MY, Sing NW, Ling KV. Augmented reality systems for medical applications. IEEE engineering in medicine and biology magazine : the quarterly magazine of the Engineering in Medicine & Biology Society. 1998;17(3):49-58. 8. Kramers M, Armstrong R, Bakhshmand SM, Fenster A, de Ribaupierre S, Eagleson R. Evaluation of a mobile augmented reality application for image guidance of

ACCEPTED MANUSCRIPT neurosurgical

interventions.

Studies in

Health

Technology &

Informatics.

2014;196(196):204. 9. Hou Y, Ma L, Zhu R. iPhone-assisted Augmented Reality Localization of Basal

RI PT

Ganglia Hypertensive Hematoma. World Neurosurgery. 2016;94:480. 10. Drummond K H, Houston T, Irvine T. The rise and fall and rise of virtual reality. Vox Media; 2014.

SC

11. Virtual reality systems. San Diego, CA: Academic Press Limited; 1993.

M AN U

12. Roberts DW, Strohbehn JW, Heach JF. A frameless stereotactic integration of computeried tomographic image and operating microscope. J Neurosurgery. 1986;65:545-549

13. Feiner SK. Augmented reality: a new way of seeing. Scientific American.

TE D

2002;286(4):48-55.

14. Golfinos JG, Fitzpatrick BC, Smith LR. Clinical use of a frameless stereotactic arm: results of 325 cases. Journal of Neurosurgery. 1995;83(2):197-205.

EP

15. Olson JJ, Shepherd S, Bakay RA. The EasyGuide Neuro image-guided surgery

AC C

system. Neurosurgery. 1997;40(5):1092. 16. Villalobos H, Germano IM. Clinical evaluation of multimodality registration in frameless stereotaxy. Computer Aided Surgery. 1999;4(1):45-49. 17. Mascott CR, Sol JC, Bousquet P, Lagarrigue J, Lzaorthes Y, Lauwers-Cances V. Quantification of true in vivo (application) accuracy in cranial image-guided surgery: influence of mode of patient registration. Neurosurgery. 2006;59(1):146-56. 18. Da SEJ, Leal AG, Milano JB, Da SLJ, Clemente RS, Ramina R. Image-guided

ACCEPTED MANUSCRIPT surgical planning using anatomical landmarks in the retrosigmoid approach. Acta Neurochirurgica. 2010;152(5):905-910. 19. Wolfsberger S, Rössler K, Regatschnig R, Ungersbock K. Anatomical landmarks

RI PT

for image registration in frameless stereotactic neuronavigation. Neurosurgical Review. 2002; 25(1-2):68-72.

20. Sipos EP, Tebo SA, Zinreich SJ, Long DM, Brem H. In vivo accuracy testing and

SC

clinical experience with the ISG Viewing Wand. Neurosurgery. 1996;39(1):194-204.

M AN U

21. Fan X, Roberts DW, Schaewe TJ, Ji S, Holton LH, Simon DA, et al.Intraoperative image updating for brain shift following dural opening. Journal of Neurosurgery. 2016;1.

22. Sun GC, Chen XL, Hou YZ, Yu XG, Ma XD, Liu G. Image-guided endoscopic

TE D

surgery for spontaneous supratentorial intracerebral hematoma. 2016;1-6. 23. Eftekhar B. App-assisted external ventricular drain insertion. Journal of Neurosurgery. 2015;1.

EP

24. Hou YZ, Ma LC, Zhu RY, Chen XL, Zhang J. A Low-Cost iPhone-Assisted

AC C

Augmented Reality Solution for the Localization of Intracranial Lesions. Plos One. 2016;11(7):e0159185.

Figure legends

Axial, coronal, sagittal MRI image demonstrated a superficial supratentorial meningioma (Fig. A, B, C). The superior view of the reconstructed model revealed the relationship among the following structures: the lesion (yellow), marker (red), center

ACCEPTED MANUSCRIPT of the lesion (dark blue), sagittal sinus with cortical vein (light blue) (Fig. D). Superior view of the patient’s head with marker (Vitamin E capsule) before surgery (Fig. E). Screen capture of the Sina during registration (Fig. F). A U-shaped skin

RI PT

incision was marked according to the location of lesion (Fig. G). Intraoperative photography verified the location of lesion (Fig. H) and the lesion was removed (Fig.

AC C

EP

TE D

M AN U

SC

I).

ACCEPTED MANUSCRIPT Table: Data of patients undergoing operations

Age (year) / Sex

Site

47/M 65/F

Size (mm)

Pathology

Left parietal

63

Right frontal

Time required for registration (s)

Distance difference (mm)

Second

Third

Total

First

Second

Third

Average

Meningioma

79

75

70

224

6.5

6.3

6.2

6.3

27

Meningioma

71

61

66

198

3.1

3.4

3.3

3.3

54

Meningioma

69

58

56

183

4.5

4.8

4.8

4.7

23

Meningioma

58

54

63

175

4.4

4.2

4.4

4.3

15

Meningioma

55

61

52

168

4.6

4.8

4.8

4.7

74/M

Left parietal

54

Glioma

49

43

55

147

3.6

3.6

3.9

3.7

68/M

Right frontal

42

Meningioma

54

42

59

48/M

Right parietal

51

Meningioma

49

45

39

63/F

Left frontal

68

Meningioma

62/M

Left temporal

35

Meningioma

51/M

Left frontal

44

Meningioma

24

Meningioma

28

Meningioma

43/F

53/F 58/F

Right parietal Left frontal and parietal

52/F

Left frontal

20

53/F

Left frontal

27

67/F

Left frontal

25

SC

61/M

155

2.2

2.3

2.5

2.3

133

4.1

4.3

4.4

4.3

44

42

40

126

5.2

5.4

5.3

5.3

39

43

40

122

6

6.2

5.9

6.0

39

35

43

117

2.5

2.6

2.9

2.7

43

47

41

131

5.9

5.5

5.8

5.7

37

33

38

108

5.5

5.3

5.2

5.3

Meningioma

36

31

30

97

3.8

3.9

3.5

3.7

Meningioma

29

33

32

94

3.9

4.1

4.1

4.0

Meningioma

27

30

32

89

4.5

4.8

4.8

4.7

TE D

54/M

RI PT

Right parietal Right frontal and parietal Right parietal

M AN U

First

AC C

EP

M: male; F: female; Distance difference: distance distance between two methods of localization; s:second; mm: milimeter

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT Augmented reality (AR) for neuronavagation using mobile devices is presented 3D reconstruction guaranteed the accuracy of this mobile AR

AC C

EP

TE D

M AN U

SC

RI PT

This mobile AR proves to be practical and reliable

ACCEPTED MANUSCRIPT Two-dimensional

3D

Three-dimensional

AR

Augmented Reality

MRI

Magnetic Resonance Imaging

CT

Computed Tomography

DICOM

Digital Imaging and Communications in Medicine

SD

Standard Deviation

AC C

EP

TE D

M AN U

SC

RI PT

2D