- Email: [email protected]
Evaluation of tissue harmonic imaging for the diagnosis of focal liver lesions
Ultrasound in Med. & Biol., Vol. 26, No. 2, pp. 183–187, 2000 Copyright © 2000 World Federation for Ultrasound in Medicine & Biology Printed in the USA. All rights reserved 0301-5629/00/$–see front matter
PII S0301-5629(99)00134-9
● Original Contribution EVALUATION OF TISSUE HARMONIC IMAGING FOR THE DIAGNOSIS OF FOCAL LIVER LESIONS SACHIKO TANAKA, OSAMU OSHIKAWA, TERUKO SASAKI, TATSUYA IOKA and HIDEAKI TSUKUMA The Osaka Medical Center for Cancer and CVD, 1-3-3, Nakamichi, Higashinari-ku, Osaka 537-8511 Japan (Received 30 April 1999; in final form 27 August 1999)
Abstract—This was a prospective study to evaluate tissue harmonic imaging (THI) for the diagnosis of focal liver lesions. A total of 15 reviewers read 100 randomly arranged liver images, a fundamental grey-scale image (FGI) and a THI (transmitted: 2 MHz, received: 4 MHz) of each of 50 patients (29 with liver cirrhosis, 42 with focal lesions) taken from the same section. The mean value of overall accuracy for detecting lesions (presence or absence) was significantly higher with THI (82.3%) than with FGI (79.6%) (t ⴝ 1.96, p < 0.05). When only the 29 cirrhosis patients were analyzed, the difference was more significant (t ⴝ 2.48, p < 0.02). The correct count rate of the number of focal lesions was higher with THI (78.0%) than with FGI (67.0%) (t ⴝ 3.61, p < 0.005) in 23 cirrhosis patients with focal lesions. The correct diagnosis of HCC was achieved at a higher rate with THI (42.5%) than with FGI (36.8%). THI was statistically effective for detecting focal lesions, particularly in cirrhotic livers. © 2000 World Federation for Ultrasound in Medicine & Biology. Key Words: Tissue harmonic imaging, Hepatocellular carcinoma, Liver cirrhosis, Liver neoplasm, Diagnostic accuracy.
to– clutter ratio and reduces the near field artefacts (Ward et al. 1997; Averkiou et al. 1997). In the clinical use of this technique, the penetration, detail and total image clarity have been reported to be improved (Shapiro et al. 1998; Tranquart et al. 1999). However, the diagnostic value of this technique has not yet been proven. In this prospective study, the effectiveness of tissue harmonic imaging for the detection and characterization of liver tumors, particularly in the cirrhotic liver, was objectively evaluated.
INTRODUCTION AND LITERATURE The accuracy of sonography for the diagnosis of hepatocellular carcinoma (HCC) is considered to be high (Tanaka et al. 1986; Tanaka et al. 1989). Also, periodic checkup by sonography for early diagnosis of HCC in patients with hepatitis B or C viral infection was reported to be effective (Tanaka et al. 1990). However, in patients with viral liver cirrhosis or chronic viral hepatitis, the echo pattern of the liver parenchyma is rough and nodular. As a result, detection of a small malignant nodule among multiple regenerative nodules is very difficult (Kitamura et al. 1995). This represents a limitation of sonography for the early detection of HCC. A new technique, tissue harmonic imaging (THI), has been introduced (Christopher 1997; Ward et al. 1997; Averkiou et al. 1997). It is a sonographic imaging technique in which echoes from a tissue are received at a second harmonic frequency that is twice as high as the transmitted frequency. It has been reported that this new technique improves the lateral resolution and the signal-
SUBJECTS AND METHODS The equipment used was an HDI-3000 (ATL Ultrasound, Bothell, WA) ultrasound (US) scanner with a 4 –2-MHz curved array probe. Fundamental grey-scale imaging (FGI) was performed at the dynamic range of 50 dB or 55 dB, and THI was performed at the dynamic range of 50 dB on the research version setting. With this software unit, a broad spectrum beam around 2 MHz is transmitted and only a narrow-band spectrum beam near 4 MHz is received. The study subjects were 50 patients for which dynamic study of the liver by x-ray computerized tomography (CT) and/or magnetic resonance imaging (MRI)
Address correspondence to: Sachiko Tanaka, M.D., The Osaka Medical Center for Cancer & CVD, 1-3-3, Nakamichi, Higashinari-ku, Osaka, 537-8511, Japan. E-mail: [email protected] 183
184
Ultrasound in Medicine and Biology
had been performed within 1 month before sonography, and sonographic examinations were performed consecutively by the first author of this paper, during the period from April to December 1997. Oral informed consent was obtained from all the patients in accordance with the rules of the ethics committee of our institution. These 50 patients (36 men and 14 women; age range: 48 to 77 y; mean: 62.2 y) consisted of 19 with HCC, 10 with metastatic liver cancer, 7 with hemangioma, 2 with focal nodular hyperplasia (FNH), 2 with large regenerative nodules (LRN), 2 with focal fatty liver and 8 without any focal lesion (3 with hepatitis B, 3 with hepatitis C viral liver cirrhosis, and 2 with a healthy liver). A total of 16 HCCs, 1 hemangioma and 1 FNH patient also with hepatitis C viral liver cirrhosis, and 3 HCCs and 2 LRNs also had hepatitis B viral liver cirrhosis. A total of 19 HCCs, 10 metastatic tumors, 2 FNHs and 2 patients with LRN were histologically diagnosed after surgical resection (20 patients) or US-guided biopsy (13 patients). The remaining patient diagnoses were confirmed by MRI and/or x-ray CT with dynamic study. For these 50 patients, both FGI and THI liver images were taken by the first author, for the same section, by adjusting for the position and the shape of the vessels or organs, using the same depth range, zoom range and focal points. The images of the 42 patients with focal lesions revealed 1 focal lesion in 37 patients, 2 lesions in 3 patients and 3 lesions in 2 patients. The size of the focal lesions ranged from 0.9 cm to 8 cm (median: 2 cm, mean ⫾ 2 SD: 2.2 ⫾ 1.5 cm). There was no discrepancy between the sonographic and x-ray CT and/or MRI findings with regard to the position or number of focal lesions. Specialists (8 gastroenterologists and 7 sonographers) participated in this trial as reviewers. They have been engaged in routine abdominal sonographic examination, but none of them had experience with THI. The length of their experience was less than 5 y for 5 specialists, 6 –10 y for 5 others and more than 10 y for the remaining 5. In our hospital, sonographers also perform routine sonographic examination and make reports after examination, but the sonographic diagnosis is made by a gastroenterologist who reviews the images afterward. A total of 100 images (50 FGIs and 50 THIs) were randomly arranged and read by the reviewers. Only 15 s or less was allowed for evaluation of each image to simulate the instant judgement performed in real-time examinations. The reviewers then estimated the number of focal lesions, if present, and the most likely single diagnosis of the tumor, if possible. Differential diagnosis of the tumor was made according to the criteria for liver cancer, based on the sonographic pattern of the tumors (Japan Society of Ultrasonics in Medicine 1994). The reviewers were blinded to the pathological diagnosis and
Volume 26, Number 2, 2000
Table 1. Overall accuracy for detection of focal lesions (presence or absence) Mean value of the 15 reviewers (%) All cases (n ⫽ 50) LC (n ⫽ 29)
FGI THI FGI THI
79.6 82.8 77.9 84.1
Statistical analysis t p t p
⫽ ⬍ ⫽ ⬍
1.96 0.05 2.48 0.02
FGI ⫽ fundamental grey-scale imaging; THI ⫽ tissue harmonic imaging; LC ⫽ liver cirrhosis.
to each others’ findings. The doctor who prepared the images did not participate in the review. The answers of all reviewers for all 100 images were computed. The detectability of focal lesions with both FGI and THI was statistically analyzed by the paired t-test. The ability to characterize focal lesions and the influence of the reviewer’s background were also examined. RESULTS Detection of focal lesions (presence or absence) The mean values of the overall accuracy for detection of focal lesions (presence or absence) of the 15 reviewers are shown in Table 1. The mean value was higher with THI (82.8%) than with FGI (79.6%). Statistical analysis was performed of the values of overall accuracy by each reviewer with FGI and THI as the related variables (paired t-test). For all 50 patients, including 42 with focal lesions, the difference between the accuracy with THI and that with FGI was statistically significant (t ⫽ 1.96, p ⬍ 0.05). When only the 29 liver cirrhosis patients (including 23 with focal lesions) were concerned, the mean overall accuracy with THI (84.1%) was significantly higher than with FGI (77.9%) (t ⫽ 2.48, p ⬍ 0.02).
Table 2. Correct count and erroneous count rate for the number of focal lesions Correct count Under-count Over-count (%) (%) (%) All focal lesion Patients (n ⫽ 42) Focal lesions with LC (n ⫽ 23)
Statistical analysis
FGI
70.5
25.1
4.4
t ⫽ 2.04
THI FGI
75.4 67.0
18.9 27.8
5.7 5.2
p ⬍ 0.1 t ⫽ 3.605
THI
78.0
17.4
4.6
p ⬍ 0.005
FGI ⫽ fundamental grey-scale imaging; THI ⫽ tissue harmonic imaging; LC ⫽ liver cirrhosis.
Tissue harmonics for liver tumor ● S. TANAKA et al.
185
Table 3. Characterization of focal lesions, % correctly diagnosed by the reviewers
FGI THI
All patients with lesions (n ⫽ 42)
HCC (%) (n ⫽ 19)
Metastasis (%) (n ⫽ 10)
Hemangioma (n ⫽ 7)
28.6 29.2
36.8 42.5
8.0 4.0
35.2 36.2
FGI ⫽ fundamental grey-scale imaging; THI ⫽ tissue harmonic imaging; HCC ⫽ hepatocellular carcinoma.
rate of correct count with THI was related to the low incidence of under-count, as shown in Fig. 1. Statistical analysis was performed on the number of times in which correct count was made by each reviewer with FGI and THI as the related variables (paired t-test). When all 42 focal lesions were analyzed, the difference between the results with FGI and THI was not significant (t ⫽ 2.04, p ⬍ 0.1). In contrast, when only the 23 liver cirrhosis patients with focal lesions were considered, the correct count rate with THI was significantly higher than with FGI (t ⫽ 3.61, p ⬍ 0.005).
Fig. 1. Three nodules of HCC in the cirrhotic liver, confirmed by sonographic-guided aspiration biopsy. Comparing with (A) FGI, (B) THI is more contrasty and reveals 3 nodular lesions (arrow heads) more clearly. With FGI, 11 reviewers could detect only 1 lesion, and the remaining 4 reviewers could not detect any focal lesions. With THI, 6 of the 15 reviewers (40%) could detect all 3 lesions, and the other 4 (27%) could detect 2 of the 3 lesions.
Accuracy of counting of the number of focal lesions One, two or three focal lesions were visualized on the sonographic images of 42 patients. Table 2 computes the data regarding the correct count and erroneous count rates of the focal lesions with FGI and THI. With FGI, out of 630 readings (42 images read by 15 reviewers), a correct count was given 444 times (70.5%), under-count occurred 158 times (25.1%) and over-count occurred 28 times (4.4%). With THI, the rates were 75.4%, 18.9% and 5.7%, respectively. When only the 23 liver cirrhosis patients with focal lesions were the subjects, the rates of correct, underand over-count were 67.0%, 27.8% and 5.2% with FGI, and 78.0%, 17.4% and 4.6% with THI. The high
Characterization of focal lesions Characterization of the focal lesion (Table 3) was made by the reviewers in 70.7% of the 42 focal lesion patients with FGI, and in 70.4% with THI. The correct diagnosis was made in 28.6% (180) of 630 evaluations of the 42 patients on the basis of FGI and 29.2% (184) of 630 on the basis of THI. When only 19 HCC patients were used as the subjects (Figs. 2 and 3), the correct diagnosis of HCC was made 36.8% (105) of 285 patients with FGI and 42.5% (121) of 285) with THI. In 10 metastatic cancer patients, the rate of correct diagnosis was low with both imaging technique (FGI: 8.0%, THI: 4.0%) and, for 7 hemangiomas, no remarkable difference was observed (FGI: 35.2%, THI: 36.2%). Influence of the reviewer’s background The data on the influence of the reviewers’ background are shown in Table 4. The correct count rate was considered to be independent of the reviewer’s length of experience. On the other hand, regarding the differential diagnosis, the experienced group gave a correct diagnosis of HCC at a higher rate. All three groups achieved better results with THI than with FGI. DISCUSSION AND SUMMARY Harmonic imaging was first applied to achieve elimination of color Doppler motion artefacts when using intravenous contrast agents (Burns et al. 1992). Later, harmonic imaging of uncontrasted human tissue was
186
Ultrasound in Medicine and Biology
Volume 26, Number 2, 2000
Fig. 3. Encapsulated nodular HCC, 60 mm in diameter, with liver cirrhosis, was surgically resected. Compared with (A) FGI, (B) THI reveals the marginal hypoechoic zone and mosaic pattern within the nodule more clearly, suggesting an encapsulated nodular HCC (arrowheads). The number of reviewers who could give the correct answer of HCC was 11 (73%) with FGI, and 12 (80%) with THI.
introduced, and was reported to improve the signal-to– noise ratio (Christopher 1997; Ward et al. 1997). Averkiou et al. (1997) reported that the harmonic beams were narrower than the originally transmitted beams and had lower side lobes with less aberration. They also
Table 4. Influence of the reviewer’s background Fig. 2. Well-differentiated HCC, 18 mm in diameter, with fatty metamorphosis. Comparing with (A) FGI, (B) THI reveals the complex of hypoechoic area and hyperechoic area within the nodule well, which suggests a partial fatty metamorphosis within an HCC nodule (Tanaka et al. 1983). The number of reviewers who could give the correct answer of HCC was only 1 of 15 reviewers (7%) with FGI, and 5 (33%) with THI. Histological specimen of (C) the surgically resected tumor revealed a well-differentiated HCC with fatty metamorphosis.
Correct count Reviewer’s experience ⬍5y 5–10 y ⬎ 10 y
Correct diagnosis of HCC
FGI (%)
THI (%)
FGI (%)
THI (%)
70.0 70.0 71.4
74.8 75.7 75.2
25.3 37.2 47.9
33.6 40.2 53.7
FGI ⫽ fundamental grey-scale imaging; THI ⫽ tissue harmonic imaging; HCC ⫽ hepatocellular carcinoma.
Tissue harmonics for liver tumor ● S. TANAKA et al.
reported that images formed with this technique offered better border definition and enhanced tissue contrast. Regarding the clinical use of THI, Desser et al. (1998) reported that THI improved the image quality in obese or large-bodied patients. Improvement of the images of the pancreas and other organs has also been reported (Shapiro et al. 1998; Tranquart et al. 1999). However, to date, in prospective studies there have been no reports on the diagnostic value of THI. In the present study, the ability of THI to detect focal liver lesions was evaluated in comparison with FGI by performing statistical analysis of the results of 15 reviewers. This analysis demonstrated the superiority of THI to FGI for the detection of focal lesions, particularly in patients with viral liver cirrhosis. In the cirrhotic liver, proliferation of interlobular fibrotic tissue, pseudo lobule formation or regenerative nodules can be observed microscopically. This diseased tissue causes disturbance of the sound beam and results in a rough echo texture in the liver sonogram. Discrimination of a small malignant lesion within this rough and nodular liver pattern is very difficult (Kitamura et al. 1995). In addition, these cirrhotic livers are at high risk for HCC. With THI, the grey-scale contrast between small HCC and fibrotic liver tissue was more clearly distinguished than with FGI, and tumors could be readily discriminated, as shown in Fig. 1. This resulted in a lower incidence of under-count and a higher rate of correct count of the number of focal lesions. Moreover, in some HCC patients, the characteristic sonographic pattern of HCC was more clearly visualized by THI than by FGI, as shown in Figs. 2 and 3, and some reviewers were able to make the correct diagnosis of HCC only with THI. This resulted in a rather higher rate of correct diagnosis of HCC with THI. The low capability for diagnosis of metastatic tumors was possibly because of the variety of sonographic patterns and the lack of information about the original malignant focus. This was the first experience of reading THI for all the reviewers. Most of them had the impression of something curious or glittering on THI. Therefore, after they become familiar with THI, or some improvement is made in the image formation, the accuracy with THI may increase even in patients with noncirrhotic livers. The information was limited to only one frozen image and,
187
thus, the diagnostic accuracy in the present study can be presumed to be low compared with that of routine examination using real-time images. In conclusion, THI was statistically more effective than FGI for the detection of focal lesions, particularly in the cirrhotic liver. The effectiveness of THI was independent of the reviewer’s career in sonography. Therefore, the use of THI for patients with viral liver cirrhosis, a group of patients at high risk for HCC, can be expected to improve the accuracy of early diagnosis of HCC. Acknowledgements—This study was supported in part by a grant from the Foundation for Promotion of Cancer Research. The authors thank H. Tanaka, M.D., for kind advice on the statistical analysis, Tsugio Kitamra, M.D., Fumi Yoshioka, M.D., Hiroshi Kasugai, M.D., Atsuo Inoue, M.D., Hiroyuki Uehara, M.D., Takashi Murakami, M.D., K. Yamamoto, M.T., T. Hori, M.T., S. Fujikawa, M.T., S. Nishikawa, M.T., M. Nakao, M.T., and J. Fukuda, M.T., for participation as reviewers.
REFERENCES Averkiou MA, Roundhill DN, Powers JE. A new imaging technique based on the nonlinear properties of tissues. Proc IEEE Ultrason Sympos 1997;2:1561–1566. Burns PN, Powers JE, Fritzsch T. Harmonic imaging: A new imaging and Doppler method for contrast enhanced ultrasound. (Abstr.) Radiology 1992;185:142. Christopher T. Finite amplitude distortion-based inhomogeneous pulse echo ultrasonic imaging. IEEE Trans Ultrason 1997;44:125–139. Desser TS, Lane MJ, Jeffrey B. Harmonic assist study of abdominal pathology. Diagn Imaging 1998;(Suppl.):AU8 –AU10. Japan Society of Ultrasonics in Medicine. Diagnostic criteria for liver tumors. Diagnostic ultrasound: principles and clinical applications. Tokyo: Igakushoin, 1994:925. Kitamura S, Iisi H, Tatsuta M, et al. Liver with hypoechoic nodular pattern as a risk factor for hepatocellular carcinoma. Gastroenterology 1995;108:1778 –1784. Shapiro SS, Wagreich J, Parsons RB, Stancato-Pasik A, Yeh H, Lao R. Tissue harmonic imaging sonography: Evaluation of image quality compared with conventional sonography. AJR 1998;171:1203–1206. Tanaka S, Kitamura T, Imaoka S, et al. Hepatocellular carcinoma: sonographic and histologic correlation. AJR 1983;140:701–707. Tanaka S, Kitamura T, Nakanishi K, et al. Recent advances in ultrasonographic diagnosis of hepatocellular carcinoma. Cancer 1989; 63:1313–1317. Tanaka S, Kitamura T, Nakanishi K, et al. Effectiveness of periodic checkup by ultrasonography for the early diagnosis of hepatocellular carcinoma. Cancer 1990;66:2210 –2214. Tanaka S, Kitamura T, Ohshima A, et al. Diagnostic accuracy of ultrasonography for hepatocellular carcinoma. Cancer 1986;58:344 –347. Tranquart F, Grenier N, Eder V, Pourcelot L. Clinical use of ultrasound tissue harmonic imaging. Ultrasound Med Biol 1999;25:889 – 894. Ward B, Baker AC, Humphrey VF. Nonlinear propagation applied to the improvement of resolution in diagnostic ultrasound. J Acoustic Soc Am 1997;101:143–154.
PII S0301-5629(99)00134-9
● Original Contribution EVALUATION OF TISSUE HARMONIC IMAGING FOR THE DIAGNOSIS OF FOCAL LIVER LESIONS SACHIKO TANAKA, OSAMU OSHIKAWA, TERUKO SASAKI, TATSUYA IOKA and HIDEAKI TSUKUMA The Osaka Medical Center for Cancer and CVD, 1-3-3, Nakamichi, Higashinari-ku, Osaka 537-8511 Japan (Received 30 April 1999; in final form 27 August 1999)
Abstract—This was a prospective study to evaluate tissue harmonic imaging (THI) for the diagnosis of focal liver lesions. A total of 15 reviewers read 100 randomly arranged liver images, a fundamental grey-scale image (FGI) and a THI (transmitted: 2 MHz, received: 4 MHz) of each of 50 patients (29 with liver cirrhosis, 42 with focal lesions) taken from the same section. The mean value of overall accuracy for detecting lesions (presence or absence) was significantly higher with THI (82.3%) than with FGI (79.6%) (t ⴝ 1.96, p < 0.05). When only the 29 cirrhosis patients were analyzed, the difference was more significant (t ⴝ 2.48, p < 0.02). The correct count rate of the number of focal lesions was higher with THI (78.0%) than with FGI (67.0%) (t ⴝ 3.61, p < 0.005) in 23 cirrhosis patients with focal lesions. The correct diagnosis of HCC was achieved at a higher rate with THI (42.5%) than with FGI (36.8%). THI was statistically effective for detecting focal lesions, particularly in cirrhotic livers. © 2000 World Federation for Ultrasound in Medicine & Biology. Key Words: Tissue harmonic imaging, Hepatocellular carcinoma, Liver cirrhosis, Liver neoplasm, Diagnostic accuracy.
to– clutter ratio and reduces the near field artefacts (Ward et al. 1997; Averkiou et al. 1997). In the clinical use of this technique, the penetration, detail and total image clarity have been reported to be improved (Shapiro et al. 1998; Tranquart et al. 1999). However, the diagnostic value of this technique has not yet been proven. In this prospective study, the effectiveness of tissue harmonic imaging for the detection and characterization of liver tumors, particularly in the cirrhotic liver, was objectively evaluated.
INTRODUCTION AND LITERATURE The accuracy of sonography for the diagnosis of hepatocellular carcinoma (HCC) is considered to be high (Tanaka et al. 1986; Tanaka et al. 1989). Also, periodic checkup by sonography for early diagnosis of HCC in patients with hepatitis B or C viral infection was reported to be effective (Tanaka et al. 1990). However, in patients with viral liver cirrhosis or chronic viral hepatitis, the echo pattern of the liver parenchyma is rough and nodular. As a result, detection of a small malignant nodule among multiple regenerative nodules is very difficult (Kitamura et al. 1995). This represents a limitation of sonography for the early detection of HCC. A new technique, tissue harmonic imaging (THI), has been introduced (Christopher 1997; Ward et al. 1997; Averkiou et al. 1997). It is a sonographic imaging technique in which echoes from a tissue are received at a second harmonic frequency that is twice as high as the transmitted frequency. It has been reported that this new technique improves the lateral resolution and the signal-
SUBJECTS AND METHODS The equipment used was an HDI-3000 (ATL Ultrasound, Bothell, WA) ultrasound (US) scanner with a 4 –2-MHz curved array probe. Fundamental grey-scale imaging (FGI) was performed at the dynamic range of 50 dB or 55 dB, and THI was performed at the dynamic range of 50 dB on the research version setting. With this software unit, a broad spectrum beam around 2 MHz is transmitted and only a narrow-band spectrum beam near 4 MHz is received. The study subjects were 50 patients for which dynamic study of the liver by x-ray computerized tomography (CT) and/or magnetic resonance imaging (MRI)
Address correspondence to: Sachiko Tanaka, M.D., The Osaka Medical Center for Cancer & CVD, 1-3-3, Nakamichi, Higashinari-ku, Osaka, 537-8511, Japan. E-mail: [email protected] 183
184
Ultrasound in Medicine and Biology
had been performed within 1 month before sonography, and sonographic examinations were performed consecutively by the first author of this paper, during the period from April to December 1997. Oral informed consent was obtained from all the patients in accordance with the rules of the ethics committee of our institution. These 50 patients (36 men and 14 women; age range: 48 to 77 y; mean: 62.2 y) consisted of 19 with HCC, 10 with metastatic liver cancer, 7 with hemangioma, 2 with focal nodular hyperplasia (FNH), 2 with large regenerative nodules (LRN), 2 with focal fatty liver and 8 without any focal lesion (3 with hepatitis B, 3 with hepatitis C viral liver cirrhosis, and 2 with a healthy liver). A total of 16 HCCs, 1 hemangioma and 1 FNH patient also with hepatitis C viral liver cirrhosis, and 3 HCCs and 2 LRNs also had hepatitis B viral liver cirrhosis. A total of 19 HCCs, 10 metastatic tumors, 2 FNHs and 2 patients with LRN were histologically diagnosed after surgical resection (20 patients) or US-guided biopsy (13 patients). The remaining patient diagnoses were confirmed by MRI and/or x-ray CT with dynamic study. For these 50 patients, both FGI and THI liver images were taken by the first author, for the same section, by adjusting for the position and the shape of the vessels or organs, using the same depth range, zoom range and focal points. The images of the 42 patients with focal lesions revealed 1 focal lesion in 37 patients, 2 lesions in 3 patients and 3 lesions in 2 patients. The size of the focal lesions ranged from 0.9 cm to 8 cm (median: 2 cm, mean ⫾ 2 SD: 2.2 ⫾ 1.5 cm). There was no discrepancy between the sonographic and x-ray CT and/or MRI findings with regard to the position or number of focal lesions. Specialists (8 gastroenterologists and 7 sonographers) participated in this trial as reviewers. They have been engaged in routine abdominal sonographic examination, but none of them had experience with THI. The length of their experience was less than 5 y for 5 specialists, 6 –10 y for 5 others and more than 10 y for the remaining 5. In our hospital, sonographers also perform routine sonographic examination and make reports after examination, but the sonographic diagnosis is made by a gastroenterologist who reviews the images afterward. A total of 100 images (50 FGIs and 50 THIs) were randomly arranged and read by the reviewers. Only 15 s or less was allowed for evaluation of each image to simulate the instant judgement performed in real-time examinations. The reviewers then estimated the number of focal lesions, if present, and the most likely single diagnosis of the tumor, if possible. Differential diagnosis of the tumor was made according to the criteria for liver cancer, based on the sonographic pattern of the tumors (Japan Society of Ultrasonics in Medicine 1994). The reviewers were blinded to the pathological diagnosis and
Volume 26, Number 2, 2000
Table 1. Overall accuracy for detection of focal lesions (presence or absence) Mean value of the 15 reviewers (%) All cases (n ⫽ 50) LC (n ⫽ 29)
FGI THI FGI THI
79.6 82.8 77.9 84.1
Statistical analysis t p t p
⫽ ⬍ ⫽ ⬍
1.96 0.05 2.48 0.02
FGI ⫽ fundamental grey-scale imaging; THI ⫽ tissue harmonic imaging; LC ⫽ liver cirrhosis.
to each others’ findings. The doctor who prepared the images did not participate in the review. The answers of all reviewers for all 100 images were computed. The detectability of focal lesions with both FGI and THI was statistically analyzed by the paired t-test. The ability to characterize focal lesions and the influence of the reviewer’s background were also examined. RESULTS Detection of focal lesions (presence or absence) The mean values of the overall accuracy for detection of focal lesions (presence or absence) of the 15 reviewers are shown in Table 1. The mean value was higher with THI (82.8%) than with FGI (79.6%). Statistical analysis was performed of the values of overall accuracy by each reviewer with FGI and THI as the related variables (paired t-test). For all 50 patients, including 42 with focal lesions, the difference between the accuracy with THI and that with FGI was statistically significant (t ⫽ 1.96, p ⬍ 0.05). When only the 29 liver cirrhosis patients (including 23 with focal lesions) were concerned, the mean overall accuracy with THI (84.1%) was significantly higher than with FGI (77.9%) (t ⫽ 2.48, p ⬍ 0.02).
Table 2. Correct count and erroneous count rate for the number of focal lesions Correct count Under-count Over-count (%) (%) (%) All focal lesion Patients (n ⫽ 42) Focal lesions with LC (n ⫽ 23)
Statistical analysis
FGI
70.5
25.1
4.4
t ⫽ 2.04
THI FGI
75.4 67.0
18.9 27.8
5.7 5.2
p ⬍ 0.1 t ⫽ 3.605
THI
78.0
17.4
4.6
p ⬍ 0.005
FGI ⫽ fundamental grey-scale imaging; THI ⫽ tissue harmonic imaging; LC ⫽ liver cirrhosis.
Tissue harmonics for liver tumor ● S. TANAKA et al.
185
Table 3. Characterization of focal lesions, % correctly diagnosed by the reviewers
FGI THI
All patients with lesions (n ⫽ 42)
HCC (%) (n ⫽ 19)
Metastasis (%) (n ⫽ 10)
Hemangioma (n ⫽ 7)
28.6 29.2
36.8 42.5
8.0 4.0
35.2 36.2
FGI ⫽ fundamental grey-scale imaging; THI ⫽ tissue harmonic imaging; HCC ⫽ hepatocellular carcinoma.
rate of correct count with THI was related to the low incidence of under-count, as shown in Fig. 1. Statistical analysis was performed on the number of times in which correct count was made by each reviewer with FGI and THI as the related variables (paired t-test). When all 42 focal lesions were analyzed, the difference between the results with FGI and THI was not significant (t ⫽ 2.04, p ⬍ 0.1). In contrast, when only the 23 liver cirrhosis patients with focal lesions were considered, the correct count rate with THI was significantly higher than with FGI (t ⫽ 3.61, p ⬍ 0.005).
Fig. 1. Three nodules of HCC in the cirrhotic liver, confirmed by sonographic-guided aspiration biopsy. Comparing with (A) FGI, (B) THI is more contrasty and reveals 3 nodular lesions (arrow heads) more clearly. With FGI, 11 reviewers could detect only 1 lesion, and the remaining 4 reviewers could not detect any focal lesions. With THI, 6 of the 15 reviewers (40%) could detect all 3 lesions, and the other 4 (27%) could detect 2 of the 3 lesions.
Accuracy of counting of the number of focal lesions One, two or three focal lesions were visualized on the sonographic images of 42 patients. Table 2 computes the data regarding the correct count and erroneous count rates of the focal lesions with FGI and THI. With FGI, out of 630 readings (42 images read by 15 reviewers), a correct count was given 444 times (70.5%), under-count occurred 158 times (25.1%) and over-count occurred 28 times (4.4%). With THI, the rates were 75.4%, 18.9% and 5.7%, respectively. When only the 23 liver cirrhosis patients with focal lesions were the subjects, the rates of correct, underand over-count were 67.0%, 27.8% and 5.2% with FGI, and 78.0%, 17.4% and 4.6% with THI. The high
Characterization of focal lesions Characterization of the focal lesion (Table 3) was made by the reviewers in 70.7% of the 42 focal lesion patients with FGI, and in 70.4% with THI. The correct diagnosis was made in 28.6% (180) of 630 evaluations of the 42 patients on the basis of FGI and 29.2% (184) of 630 on the basis of THI. When only 19 HCC patients were used as the subjects (Figs. 2 and 3), the correct diagnosis of HCC was made 36.8% (105) of 285 patients with FGI and 42.5% (121) of 285) with THI. In 10 metastatic cancer patients, the rate of correct diagnosis was low with both imaging technique (FGI: 8.0%, THI: 4.0%) and, for 7 hemangiomas, no remarkable difference was observed (FGI: 35.2%, THI: 36.2%). Influence of the reviewer’s background The data on the influence of the reviewers’ background are shown in Table 4. The correct count rate was considered to be independent of the reviewer’s length of experience. On the other hand, regarding the differential diagnosis, the experienced group gave a correct diagnosis of HCC at a higher rate. All three groups achieved better results with THI than with FGI. DISCUSSION AND SUMMARY Harmonic imaging was first applied to achieve elimination of color Doppler motion artefacts when using intravenous contrast agents (Burns et al. 1992). Later, harmonic imaging of uncontrasted human tissue was
186
Ultrasound in Medicine and Biology
Volume 26, Number 2, 2000
Fig. 3. Encapsulated nodular HCC, 60 mm in diameter, with liver cirrhosis, was surgically resected. Compared with (A) FGI, (B) THI reveals the marginal hypoechoic zone and mosaic pattern within the nodule more clearly, suggesting an encapsulated nodular HCC (arrowheads). The number of reviewers who could give the correct answer of HCC was 11 (73%) with FGI, and 12 (80%) with THI.
introduced, and was reported to improve the signal-to– noise ratio (Christopher 1997; Ward et al. 1997). Averkiou et al. (1997) reported that the harmonic beams were narrower than the originally transmitted beams and had lower side lobes with less aberration. They also
Table 4. Influence of the reviewer’s background Fig. 2. Well-differentiated HCC, 18 mm in diameter, with fatty metamorphosis. Comparing with (A) FGI, (B) THI reveals the complex of hypoechoic area and hyperechoic area within the nodule well, which suggests a partial fatty metamorphosis within an HCC nodule (Tanaka et al. 1983). The number of reviewers who could give the correct answer of HCC was only 1 of 15 reviewers (7%) with FGI, and 5 (33%) with THI. Histological specimen of (C) the surgically resected tumor revealed a well-differentiated HCC with fatty metamorphosis.
Correct count Reviewer’s experience ⬍5y 5–10 y ⬎ 10 y
Correct diagnosis of HCC
FGI (%)
THI (%)
FGI (%)
THI (%)
70.0 70.0 71.4
74.8 75.7 75.2
25.3 37.2 47.9
33.6 40.2 53.7
FGI ⫽ fundamental grey-scale imaging; THI ⫽ tissue harmonic imaging; HCC ⫽ hepatocellular carcinoma.
Tissue harmonics for liver tumor ● S. TANAKA et al.
reported that images formed with this technique offered better border definition and enhanced tissue contrast. Regarding the clinical use of THI, Desser et al. (1998) reported that THI improved the image quality in obese or large-bodied patients. Improvement of the images of the pancreas and other organs has also been reported (Shapiro et al. 1998; Tranquart et al. 1999). However, to date, in prospective studies there have been no reports on the diagnostic value of THI. In the present study, the ability of THI to detect focal liver lesions was evaluated in comparison with FGI by performing statistical analysis of the results of 15 reviewers. This analysis demonstrated the superiority of THI to FGI for the detection of focal lesions, particularly in patients with viral liver cirrhosis. In the cirrhotic liver, proliferation of interlobular fibrotic tissue, pseudo lobule formation or regenerative nodules can be observed microscopically. This diseased tissue causes disturbance of the sound beam and results in a rough echo texture in the liver sonogram. Discrimination of a small malignant lesion within this rough and nodular liver pattern is very difficult (Kitamura et al. 1995). In addition, these cirrhotic livers are at high risk for HCC. With THI, the grey-scale contrast between small HCC and fibrotic liver tissue was more clearly distinguished than with FGI, and tumors could be readily discriminated, as shown in Fig. 1. This resulted in a lower incidence of under-count and a higher rate of correct count of the number of focal lesions. Moreover, in some HCC patients, the characteristic sonographic pattern of HCC was more clearly visualized by THI than by FGI, as shown in Figs. 2 and 3, and some reviewers were able to make the correct diagnosis of HCC only with THI. This resulted in a rather higher rate of correct diagnosis of HCC with THI. The low capability for diagnosis of metastatic tumors was possibly because of the variety of sonographic patterns and the lack of information about the original malignant focus. This was the first experience of reading THI for all the reviewers. Most of them had the impression of something curious or glittering on THI. Therefore, after they become familiar with THI, or some improvement is made in the image formation, the accuracy with THI may increase even in patients with noncirrhotic livers. The information was limited to only one frozen image and,
187
thus, the diagnostic accuracy in the present study can be presumed to be low compared with that of routine examination using real-time images. In conclusion, THI was statistically more effective than FGI for the detection of focal lesions, particularly in the cirrhotic liver. The effectiveness of THI was independent of the reviewer’s career in sonography. Therefore, the use of THI for patients with viral liver cirrhosis, a group of patients at high risk for HCC, can be expected to improve the accuracy of early diagnosis of HCC. Acknowledgements—This study was supported in part by a grant from the Foundation for Promotion of Cancer Research. The authors thank H. Tanaka, M.D., for kind advice on the statistical analysis, Tsugio Kitamra, M.D., Fumi Yoshioka, M.D., Hiroshi Kasugai, M.D., Atsuo Inoue, M.D., Hiroyuki Uehara, M.D., Takashi Murakami, M.D., K. Yamamoto, M.T., T. Hori, M.T., S. Fujikawa, M.T., S. Nishikawa, M.T., M. Nakao, M.T., and J. Fukuda, M.T., for participation as reviewers.
REFERENCES Averkiou MA, Roundhill DN, Powers JE. A new imaging technique based on the nonlinear properties of tissues. Proc IEEE Ultrason Sympos 1997;2:1561–1566. Burns PN, Powers JE, Fritzsch T. Harmonic imaging: A new imaging and Doppler method for contrast enhanced ultrasound. (Abstr.) Radiology 1992;185:142. Christopher T. Finite amplitude distortion-based inhomogeneous pulse echo ultrasonic imaging. IEEE Trans Ultrason 1997;44:125–139. Desser TS, Lane MJ, Jeffrey B. Harmonic assist study of abdominal pathology. Diagn Imaging 1998;(Suppl.):AU8 –AU10. Japan Society of Ultrasonics in Medicine. Diagnostic criteria for liver tumors. Diagnostic ultrasound: principles and clinical applications. Tokyo: Igakushoin, 1994:925. Kitamura S, Iisi H, Tatsuta M, et al. Liver with hypoechoic nodular pattern as a risk factor for hepatocellular carcinoma. Gastroenterology 1995;108:1778 –1784. Shapiro SS, Wagreich J, Parsons RB, Stancato-Pasik A, Yeh H, Lao R. Tissue harmonic imaging sonography: Evaluation of image quality compared with conventional sonography. AJR 1998;171:1203–1206. Tanaka S, Kitamura T, Imaoka S, et al. Hepatocellular carcinoma: sonographic and histologic correlation. AJR 1983;140:701–707. Tanaka S, Kitamura T, Nakanishi K, et al. Recent advances in ultrasonographic diagnosis of hepatocellular carcinoma. Cancer 1989; 63:1313–1317. Tanaka S, Kitamura T, Nakanishi K, et al. Effectiveness of periodic checkup by ultrasonography for the early diagnosis of hepatocellular carcinoma. Cancer 1990;66:2210 –2214. Tanaka S, Kitamura T, Ohshima A, et al. Diagnostic accuracy of ultrasonography for hepatocellular carcinoma. Cancer 1986;58:344 –347. Tranquart F, Grenier N, Eder V, Pourcelot L. Clinical use of ultrasound tissue harmonic imaging. Ultrasound Med Biol 1999;25:889 – 894. Ward B, Baker AC, Humphrey VF. Nonlinear propagation applied to the improvement of resolution in diagnostic ultrasound. J Acoustic Soc Am 1997;101:143–154.