A Radiomic-based Machine Learning Algorithm to Reliably Differentiate Benign Renal Masses from Renal Cell Carcinoma

Area of Science:

• Radiology
• Oncology
• Artificial Intelligence

Background:

• Many patients undergo treatment for renal masses where active surveillance might be more appropriate.
• Distinguishing benign from malignant renal masses is crucial for appropriate patient management.

Purpose of the Study:

• To determine if radiomic-based machine learning platforms can accurately distinguish benign from malignant renal masses.
• To evaluate the diagnostic performance of these models.

Main Methods:

• Radiomic analysis of preoperative computed tomography (CT) scans was performed on 684 patients.
• Decision tree analysis identified important clinical and radiomic variables.
• Random Forest and REAL Adaboost predictive models were developed and evaluated using receiver operating characteristic (ROC) curves.

Main Results:

• The radiomic predictive models differentiated benign from malignant renal masses with an area under the curve (AUC) of 0.84.
• For small renal masses, the discriminatory AUC was 0.77.
• Supplementing negative or nondiagnostic biopsies with radiomic analysis increased accuracy from 83.3% to 93.4%.

Conclusions:

• Radiomic-based predictive modeling shows promise in distinguishing benign from malignant renal masses.
• Clinical factors did not significantly enhance diagnostic accuracy.
• Improved diagnostic predictability could enhance patient selection for surgery and increase the use of active surveillance.