Radiomics-Based Machine Learning versus the MAP Score for Predicting Haematocrit Decline after Laparoscopic Partial...

Area of Science:

• Urology
• Radiology
• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Renal cell carcinoma (RCC) is a common malignancy.
• Laparoscopic partial nephrectomy (LPN) is a standard treatment for localized RCC.
• Predicting postoperative hematocrit (Hct) reduction is crucial for patient management.

Purpose of the Study:

• To compare the predictive accuracy of the Mayo Adhesive Probability (MAP) score and radiomics-based machine learning models for postoperative Hct reduction after LPN for RCC.
• To evaluate the utility of preoperative computed tomography (CT) images in predicting Hct decline.

Main Methods:

• Retrospective observational study involving 53 RCC patients undergoing LPN.
• Patients were categorized into high (>6 units) and low Hct decline groups.
• MAP scores were assigned, and 107 radiomics features were extracted from CT images.
• Seven machine learning classifiers were trained and validated using 10-fold cross-validation.

Main Results:

• The MAP score achieved an AUC of 0.777 for predicting high Hct decline.
• The radiomics-based Naive Bayes model demonstrated superior performance with an AUC of 0.925, 95.0% accuracy, 90.0% sensitivity, and 100.0% specificity.

Conclusions:

• Radiomics analysis combined with machine learning significantly outperforms the MAP score in predicting postoperative Hct decline after LPN for RCC.
• Integrating radiomics into preoperative assessment can optimize surgical planning and enhance patient safety.
• Further multicenter prospective studies are recommended before clinical implementation.