Staging hepatic fibrosis: Computer-aided analysis of hepatic contours on gadolinium ethoxybenzyl diethylenetriaminepentaacetic acid-enhanced hepatocyte-phase magnetic resonance imaging

Authors


  • Potential conflict of interest: Nothing to report.

To the Editor:

Chronic liver diseases can lead to hepatic fibrosis, cirrhosis, the development of hepatocellular carcinoma, and contribute substantially to healthcare costs.1 Detection and grading of hepatic fibrosis currently requires a biopsy, which subjects the patient to a risk of serious complications.2 Liver surface nodularity reflects the presence of regenerative nodules and fibrous septa, which are the essential histologic findings for the diagnosis of cirrhosis.3, 4 Gadolinium ethoxybenzyl diethylenetriaminepentaacetic acid (Gd-EOB-DTPA) is a liver-specific magnetic resonance imaging (MRI) contrast medium and its hepatocyte-phase images yield excellent hepatic enhancement and the objective delineation of hepatic contour morphology. We thus conducted a computer-aided diagnosis algorithm based on the hepatic contour morphological features such as surface nodularity for predicting hepatic fibrosis stages.

Between February 2010 and January 2011, 87 patients (56 male, 31 female; age range, 39-85 years, hepatitis C in 72, hepatitis B in 12, alcohol abuse in 2, and cryptogenic in 1) with pathologically proven hepatic fibrosis stages underwent Gd-EOB-DTPA-enhanced MRI with a 3T superconducting system. Stage was determined by: hepatectomy (n = 32) to treat hepatic tumors or percutaneous liver biopsy (n = 55). Fibrosis stages were determined according to the established criteria5: F0 (n = 9); F1 (n = 16); F2 (n = 13); F3 (n = 21); and F4 (n = 28).

According to our algorithm (Fig. 1), the mean ± SD was significantly higher in patients with ≥F3 than with ≤F2 (P < 0.001, Tukey criterion). No significant difference was found among F0, F1, and F2. Although the mean ± SD of F4 was higher than that of F3 (P < 0.05), there was a considerable overlap in their distribution. Post-hoc power analysis showed that we had 80.9% power to detect a 15% difference in linear multiple regression analysis between fibrosis stages. Both the sensitivity and specificity of SD for the diagnosis of hepatic fibrosis stages ≥F3 were 100% using a cutoff value of 0.65 (Fig. 1c).

Figure 1.

A 73-year-old woman with fibrosis stage F4 due to hepatitis C virus. Axial T1-weighted image (fat-suppressed three-dimensional spoiled fast field-echo [TR/TE, 4.0/1.9 ms]; field-of-view, 42 × 29 cm; 336 × 168 image matrix [512 × 512 reconstruction]; parallel imaging factor, 1; flip angle, 13°; slice thickness, 4.4-mm section thickness with 2.2-mm overlap; acquisition time, 90 slices per each phase during 22-second breath holding) at the hepatic hilar level obtained in the hepatocyte phase (A) was magnified twice to trace the hepatic surface profile (B). More than 128 points with quarter pixel size were requested in this stage. Hepatic profile curve (f(x)) was obtained by making a straight line between the start and endpoint of the hepatic profile obtained in (B), then rotating parallel to the x-axis. An approximate curve (D(x)) was determined by a least-square approach with an n-th degree polynomial equation which produced the highest R2 value. An approximate curve (D(x)) in this case was expressed by a twenty-ninth degree polynomial equation with R2 value of 0.788. The difference between D(x) and f(x) was calculated (S(x) = D(x) − f(x)) and then a standard deviation of S(x) (SD) was used for possible MRI parameters. (C) Mean ± SD was higher in patients with F4 than with F3 (P < 0.05), whereas they still overlapped. No significant difference was found between F0, F1, and F2. ns, not significant.

We have successfully applied a computer-aided analysis of hepatic contour that was highly accurate in diagnosing hepatic fibrosis stages F3 and F4 and may be a useful imaging biomarker for staging hepatic fibrosis.

Acknowledgements

Supported in part by the Health and Labour Sciences Research Grants for Third Term Comprehensive Control Research for Cancer.

Satoshi Goshima M.D.*, Masayuki Kanematsu M.D.*, Tatsunori Kobayashi R.T., M.S.†, Takahiro Furukawa B.S.†, Xuejun Zhang Ph.D.†, Hiroshi Fujita Ph.D.†, Haruo Watanabe M.D.*, Hiroshi Kondo M.D.*, Noriyuki Moriyama M.D.‡, Kyongtae T. Bae M.D, Ph.D.§, * Department of Radiology, Gifu University Hospital, Gifu, Japan, † Department of Intelligent Image Information, Division of Regeneration and Advanced Medical Sciences, Graduate School of Medicine, Gifu University, Gifu, Japan, ‡ Department of Diagnostic Radiology, National Cancer Center Hospital, Tokyo, Japan, § Department of Radiology, University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania, USA.

Ancillary