Tumor Microenvironment, Radiology, and Artificial Intelligence: Should We Consider Tumor Periphery?

Area of Science:

• Medical Imaging
• Oncology
• Machine Learning

Background:

• The tumor microenvironment (TME) influences tumor progression and immune response.
• Accurate differentiation of benign and malignant lesions is crucial for patient management.

Purpose of the Study:

• To investigate the diagnostic potential of the TME in differentiating benign and malignant lesions.
• To utilize image quantification and machine learning for improved diagnostic accuracy.

Main Methods:

• 229 breast lesions and 220 cervical lymph nodes were analyzed.
• Radiomics features were extracted from extended contours of lesions.
• A support vector machine (SVM) classifier was developed using significant radiomics features.

Main Results:

• Extended contours yielded superior radiomics features compared to standard radiologist contours.
• The SVM model with extended contour features achieved high accuracy (AUC 0.887 for breast, 0.970 for lymph nodes).

Conclusions:

• The tumor periphery and TME are vital for medical imaging diagnosis.
• Considering the immediate tumor periphery in image quantification can enhance the differentiation of benign and malignant lesions.