Enhancing Renal Tumor Detection: Leveraging Artificial Neural Networks in Computed Tomography Analysis

Area of Science:

• Medical Imaging
• Artificial Intelligence in Oncology
• Renal Cell Carcinoma Research

Background:

• Renal cell carcinoma (RCC) is a prevalent European cancer, with significant overdiagnosis rates (11-30.9%) from radiological studies before surgery.
• Accurate differentiation between malignant and benign renal tumors is crucial for effective patient management and avoiding unnecessary interventions.

Purpose of the Study:

• To develop an artificial neural network (ANN) model utilizing computed tomography (CT) images.
• To enhance the distinction between malignant and benign renal tumors as a supplementary tool for clinical decision-making and active surveillance.

Main Methods:

• A retrospective analysis of 357 renal tumor cases using axial CT images.
• Training an artificial neural network (ANN) using the VGG-16 architecture on 7207 arterial-phase CT images.
• Validation of the ANN model on a test dataset of 38 cases (10 benign, 28 malignant) with subgroup randomization.

Main Results:

• The trained ANN achieved an overall accuracy of 81.6% in classifying renal tumors.
• Specific performance metrics included 82.1% sensitivity, 80.0% specificity, and an F1 score of 86.8%.
• The ANN correctly classified 23 of 28 malignant tumors and 8 of 10 benign tumors in the validation set.

Conclusions:

• The developed artificial neural network (ANN) demonstrates significant potential in differentiating benign from malignant renal tumors.
• This AI-driven approach shows promise as an adjunctive tool in radiological assessments for renal masses, supporting active surveillance strategies.