Are radiomics features universally applicable to different organs?

Area of Science:

• Radiomics and Medical Imaging Analysis
• Oncology and Tumor Microenvironment Research
• Biostatistics and Survival Analysis

Background:

• Radiomics features identify tumor characteristics and microenvironment.
• Growing interest in applying radiomics across different organs.
• Exploration of common radiomics features in diverse organ environments.

Purpose of the Study:

• To investigate the generalizability of radiomics features across different organs.
• To determine if radiomics models developed for one organ can predict survival in others.
• To identify common radiomics features applicable to multiple organs.

Main Methods:

• Analysis of four datasets across three organs (lungs, kidneys, brains).
• Construction and evaluation of a radiomics score model trained on lung data.
• Assessment of the model's ability to stratify risk in independent lung, kidney, and brain tumor datasets.

Main Results:

• Organ-level analysis revealed distinct histogram patterns and parameters.
• The lung-trained radiomics score effectively stratified survival only in lung data, not in kidney or brain data.
• No common specific features were found between training and test sets, but a common feature category (texture) was identified.

Conclusions:

• Radiomics score models for survival prediction are predominantly organ-specific.
• Application of radiomics models to different organs necessitates careful consideration of organ-specific properties.
• The general applicability of radiomics models across organs remains limited without organ-specific adaptations.