Determination of liver volume from CT scans using histogram cluster analysis

Area of Science:

• Medical Imaging
• Computational Biology
• Radiology

Background:

• Accurate liver volume measurement is crucial for clinical decisions.
• Manual methods are time-consuming and prone to variability.
• Novel computational approaches are needed for precise in vivo organ volumetry.

Purpose of the Study:

• To evaluate a semiautomatic algorithm for total liver volume calculation using the Histogram Cluster Analysis Procedure (HICAP).
• To compare the algorithm's performance against a human observer.
• To assess the repeatability and clinical feasibility of the HICAP-based method.

Main Methods:

• Developed a semiautomatic computer algorithm incorporating HICAP for liver volume calculation from CT scans.
• Compared algorithm-derived volumes with manual measurements by an experienced radiologist.
• Assessed day-to-day variability by repeating calculations over 3-12 months.
• Explored HICAP's bivariate mode using nonenhanced and contrast-enhanced CT scans.

Main Results:

• The HICAP algorithm showed a median absolute difference of 3.6% compared to manual measurements (r2 = 0.99).
• Median day-to-day variability of computer-calculated volumes was 1.9%.
• Bivariate HICAP analysis achieved a significantly lower error of 0.4% compared to univariate analysis (4.1%).

Conclusions:

• The HICAP statistical clustering algorithm offers a clinically accurate and repeatable method for in vivo liver volume determination.
• The semiautomatic approach enhances feasibility and reduces inter-observer variability in liver volumetry.
• HICAP's ability to incorporate multiple imaging variables improves accuracy.