Renal Tumor Quantification and Classification in Contrast-Enhanced Abdominal CT

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Kidney cancer affects nearly a quarter-million people in the USA, with 51,000 new diagnoses annually.
• Current manual tumor size measurements are 2D, lack 3D detail, and exhibit significant operator variability.
• There is a need for improved methods to accurately assess tumor characteristics and treatment response.

Purpose of the Study:

• To develop and validate a computer-assisted radiology tool for assessing renal tumors in contrast-enhanced CT scans.
• To improve the accuracy and consistency of tumor size, volume, and enhancement quantification.
• To aid in tumor diagnosis and monitoring treatment responses for kidney cancer.

Main Methods:

• An algorithm combining anisotropic diffusion, fast-marching, and geodesic level-sets for tumor segmentation.
• A novel statistical refinement step to adapt segmentation to lesion shape.
• Quantification of 3D size, volume, and enhancement, enabling serial management over time.

Main Results:

• Robust segmentation of renal tumors was achieved.
• Semi-automated quantification showed results within the limits of inter-observer variability compared to manual measurements.
• Analysis of lesion enhancement demonstrated significant separation between different lesion types (cysts, VHL, HPRC) with p < 0.004.

Conclusions:

• The developed computer-assisted tool offers accurate 3D assessment of renal tumors.
• The method shows potential for improved disease monitoring, clinical trials, and noninvasive surveillance of kidney cancer.
• Enhanced quantification of tumor characteristics can aid in differential diagnosis and treatment response evaluation.