A Decision-Support Tool for Renal Mass Classification.

Area of Science:

• Medical imaging analysis
• Machine learning in oncology
• Radiomics and artificial intelligence

Background:

• Accurate identification of renal mass malignancy is crucial for patient management.
• Current diagnostic methods rely on radiologist interpretation, which can be subjective.
• Radiomics offers quantitative imaging features for improved diagnostic accuracy.

Purpose of the Study:

• To evaluate the effectiveness of statistical relational machine learning algorithms for renal mass malignancy identification.
• To develop human-interpretable classification models using radiomics features.
• To compare the performance of these novel algorithms against existing methods and expert radiologists.

Main Methods:

• Extraction of radiomics features (texture, signal intensity) from multiphase contrast-enhanced CT images.
• Application of relational functional gradient boosting (RFGB), a novel statistical relational machine learning technique.
• Development of human-interpretable classification models for malignancy identification.

Main Results:

• RFGB models achieved superior performance in identifying renal mass malignancy.
• The developed models demonstrated higher accuracy compared to standard machine learning approaches.
• RFGB outperformed the current gold standard of visual qualification by radiologists.

Conclusions:

• Statistical relational machine learning, particularly RFGB, is a viable and effective tool for improving renal mass malignancy diagnosis.
• Radiomics-based, interpretable AI models offer a promising advancement over subjective radiologist assessments.
• This approach has the potential to enhance diagnostic accuracy and patient care in renal oncology.