Radiomic Artificial Intelligence Models Predicting the Response of Colorectal Cancer Liver Metastases to...

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Colorectal cancer liver metastasis (CRCLM) significantly reduces patient survival.
• Chemotherapy response in CRCLM is highly variable, complicating treatment.
• AI radiomics models using pre-treatment imaging offer potential for predicting treatment response.

Purpose of the Study:

• To systematically review the literature on AI radiomics models for predicting chemotherapy response in CRCLM.
• To assess the performance and limitations of current AI radiomics models in this context.

Main Methods:

• Systematic literature search conducted on Clarivate (Web of Science) and Ovid (Embase, MEDLINE) up to April 8, 2025.
• Included studies focused on AI radiomics models predicting CRCLM response to chemotherapy, alone or with targeted therapies.
• Data extraction and risk of bias assessment performed following PRISMA guidelines.

Main Results:

• Thirteen studies met inclusion criteria, all focusing on model development and validation, with no implementation studies.
• AI models showed varied predictive performance, with six studies reporting good and three reporting excellent predictive ability.
• Significant heterogeneity existed in imaging modalities, lesion criteria, dimensionality, and use of clinical features; most models lacked external validation.

Conclusions:

• AI radiomics models demonstrate potential for predicting chemotherapy response in CRCLM.
• Clinical applicability is currently limited by insufficient external validation and methodological inconsistencies.
• Standardization of methods and prospective validation are crucial for clinical translation.