Purpose: Traditionally, liver surgery relies on Couinaud's liver segment classification. As the position and shape of these segments are variable and their borders are hidden within the homogeneous liver mass, the accuracy of segment identification methods needs computer-aided reevaluation.
Method: The segmental liver anatomy of 23 patients receiving diagnostic helical CT scans because of suspected intrahepatic lesion was analyzed with the aid of a computer-based operation-planning system. We compared the standard Couinaud classification, which depends particularly on the main stems of portal and hepatic veins, with a method that calculates the segment borders by analyzing the complete portal venous tree. Volume, shape, and position of the liver segments found by each method were compared.
Results: With reference to the portal vein-based method, segmental volumes were overestimated by the classic Couinaud method by up to 24% and underestimated to 13%. Volumes of Couinaud segments 4a, 7, and 8 were generally larger compared with those obtained by the portal vein-based method, whereas segments 3 and 6 were smaller. Gross variations were found in segments 5, 7, and 8. When shape and position were considered, poor correlation was found for five segments (median κ = 0.35–0.45). Only segments 2, 7, and 8 had κ values clearly above 0.45 in the majority of cases. The plane that divides the two hemilivers along the middle hepatic vein and the border between the left sector (segments 2 and 3) and the medial sector (4a and 4b) were found in both methods with very good conformity (κ > 0.75).
Conclusion: Couinaud's method of dividing the liver into eight autonomous liver segments has to be accepted as a good approximation. Nevertheless, the volume, position, and shape of these segments and their segmental borders show significant variability.
From the Departments of Surgery (L. Fischer, T. Lehnert, E. Klar, and W. Lamadé) and Radiology (L. Grenacher), University of Heidelberg, and Department of Medical Image Processing (C. Cardenas, M. Thorn, M. Vetter, and J.-P. Meinzer) and Central Unit Biostatistics (A. Benner), German Cancer Research Center, Heidelberg, Germany.
Address correspondence and reprint requests to Dr. L. Fischer at Department of Surgery, University of Heidelberg, INF Feld 110, 69120 Heidelberg, Germany. E-mail: Lars_Fischer@med.uni-heidelberg.de