Deep learning with a convolutional neural network model to differentiate renal parenchymal tumors: a preliminary...

Area of Science:

• Radiology
• Artificial Intelligence
• Oncology

Background:

• Accurate differentiation of renal parenchymal tumor subtypes is crucial for effective treatment planning.
• Advancements in medical imaging facilitate early detection, yet radiologic classification of tumor subtypes remains challenging.
• Deep learning offers potential solutions for enhancing diagnostic accuracy in medical image analysis.

Purpose of the Study:

• To develop and evaluate a novel deep convolutional neural network (CNN) model for classifying renal parenchymal tumor subtypes.
• To investigate the efficacy of the CNN model in identifying specific subtypes including clear cell renal cell carcinoma (ccRCC), chromophobe renal cell carcinoma (chRCC), angiomyolipoma (AML), and papillary renal cell carcinoma (pRCC).
• To assess the diagnostic performance of the CNN model using T2-weighted fat saturation sequence magnetic resonance (MR) images.

Main Methods:

• A retrospective study involving 199 patients with pathologically confirmed renal parenchymal tumors.
• T2-weighted fat saturation sequence MR images from 1.5 T or 3.0 T scanners were utilized.
• A deep learning model was constructed, and its performance was evaluated using receiver operating characteristic (ROC) curves, calculating accuracy, precision, sensitivity, specificity, F1-score, and area under the curve (AUC).

Main Results:

• The deep CNN model achieved an overall accuracy of 60.4% and an average accuracy of 61.7%.
• The macro-average area under the curve (AUC) for the model was 0.82.
• Specific AUCs for tumor subtypes were: ccRCC (0.94), chRCC (0.78), AML (0.80), and pRCC (0.76).

Conclusions:

• The developed deep CNN model demonstrates utility in classifying renal parenchymal tumor subtypes.
• The model achieved relatively high diagnostic accuracy when applied to T2-weighted fat saturation sequence MR images.
• This deep learning approach shows potential for improving the radiologic differentiation of renal tumors.