Texture analysis and machine learning algorithms accurately predict histologic grade in small (< 4 cm) clear cell...

Area of Science:

• Radiology and Medical Imaging
• Oncology
• Artificial Intelligence in Medicine

Background:

• Clear cell renal cell carcinoma (ccRCC) is the most common subtype of kidney cancer.
• Accurate histologic grading of ccRCC is crucial for treatment planning and prognosis.
• Non-invasive methods for predicting ccRCC grade are highly desirable.

Purpose of the Study:

• To predict the histologic grade of small clear cell renal cell carcinomas (ccRCCs) using texture analysis and machine learning algorithms.
• To evaluate the performance of different CT phases (noncontrast, corticomedullary, nephrographic) in predicting ccRCC grade.

Main Methods:

• Retrospective analysis of 52 noncontrast, 26 corticomedullary, and 35 nephrographic phase CTs of small ccRCCs.
• Calculation of histogram and texture features (GLC, GLRL) from segmented tumor images.
• Application of four machine learning algorithms (KNN, SVM, random forests, decision tree) with tenfold cross-validation.

Main Results:

• Corticomedullary (CM) phase histogram features showed statistically significant differences between low- and high-grade ccRCCs.
• CM histogram skewness and GLRL short run emphasis achieved the highest AUC of 0.82.
• Machine learning algorithms achieved an AUC of 0.97 using CM histogram features for grade prediction, outperforming other phases.

Conclusions:

• Machine learning algorithms utilizing corticomedullary phase histogram features can accurately predict the histologic grade of small ccRCCs.
• CM phase CT imaging provides superior data for ccRCC grade prediction compared to noncontrast and nephrographic phases.
• This approach offers a promising non-invasive tool for ccRCC grading.