Adaptive RAG-Assisted MRI Platform (ARAMP) for Brain Metastasis Detection and Reporting: A Retrospective Evaluation...

Area of Science:

• Artificial Intelligence in Medical Imaging
• Radiology and Medical Diagnostics
• Oncology and Cancer Imaging

Background:

• Brain metastases detection is crucial for cancer patient management.
• Current MRI interpretation can be subjective and time-consuming.
• AI-driven tools offer potential to enhance diagnostic accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate the Adaptive RAG Assistant MRI Platform (ARAMP) for brain metastasis diagnosis.
• To assess ARAMP's performance using quantitative radiomic features and a large language model.
• To improve diagnostic efficiency and standardization in radiological workflows.

Main Methods:

• Developed ARAMP integrating quantitative radiomics with an adaptive Retrieval-Augmented Generation (RAG) framework (GPT-4o).
• Retrospectively analyzed 2447 cancer patients' MRI scans; selected 100 with metastases and 100 controls.
• Evaluated ARAMP performance using neuroradiologists and an independent LLM (Gemini 2.0 Pro).

Main Results:

• Post-training, ARAMP achieved a 67.45% mean Inference Similarity score.
• Sensitivity for brain metastasis detection improved from 0.84 to 0.98 post-training.
• Inter-Inference Agreement among radiologists averaged 30.20% (p=0.01), indicating ARAMP's potential for standardization.

Conclusions:

• The adaptive RAG-based framework (ARAMP) shows promise in improving brain metastasis detection.
• ARAMP demonstrated enhanced reliability in identifying brain metastases.
• This AI platform may significantly improve diagnostic efficiency and standardization in radiological workflows.