MedicalRec: Medical recommender system for image classification without retraining

2026-06-09 · Source: cs.LG updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Health & Medical Research, Data Science & Analytics · Depth: Expert, extended

Summary

MedicalRec introduces a transformer-based recommender system and a new public dataset, MedicalRec-Bench, to address the challenge of selecting optimal models for medical image classification without extensive retraining. The system aims to reduce the significant computational power, energy consumption, and carbon emissions associated with trial-and-error model selection in healthcare AI. MedicalRec-Bench comprises over 5,000 records from 3,000 articles, detailing models tested across tasks like Skin Cancer, Tumour, Wound, Breast Cancer, and MRI classification. The dataset is available in four versions (MedicalRec I-IV) with varying feature counts (5 to 18). The MedicalRec model, utilizing BERT for embeddings and Softmax for scoring, achieved a maximum HitRate@100 of 75.5% in evaluations against 12 baseline models, demonstrating superior performance and statistical significance.

Key takeaway

For AI Scientists and Machine Learning Engineers developing medical image classification solutions, you should consider integrating recommender systems like MedicalRec to streamline model selection. This approach minimizes the environmental impact and computational costs associated with extensive model retraining. Utilize the publicly available MedicalRec-Bench dataset to benchmark your models or develop new recommendation strategies, focusing on metrics like HitRate@100 for optimal performance in identifying suitable classification models.

Key insights

A transformer-based recommender system and a new dataset optimize medical image classification model selection, reducing environmental impact.

Principles

• Deep learning's carbon footprint necessitates "Green AI" solutions.
• Recommender systems can mitigate trial-and-error model selection.
• Transformer architectures excel in sequential recommendation tasks.

Method

MedicalRec is a Transformer-based model that uses BERT for input data embeddings and a Softmax function in its final layer to calculate item (model) scores for sequential recommendation.

In practice

• Utilize MedicalRec-Bench dataset for medical image classification meta-analysis.
• Implement Transformer-based recommenders for model selection.
• Consider accuracy as a primary scoring metric for model recommendations.

Topics

• Medical Image Classification
• Recommender Systems
• Transformer Models
• Deep Learning Carbon Footprint
• MedicalRec-Bench Dataset
• Model Selection

Code references

• Ramin1Mousa/MedicalRec

Best for: Computer Vision Engineer, AI Scientist, Research Scientist, Machine Learning Engineer

Related on AIssential

• Learning neuroimaging models from health system-scale data — Health system-scale data enables AI foundation models to achieve superior, fair, and explainable neuroimaging diagnostics.
• Case-Aware Medical Image Classification with Multimodal Knowledge Graphs and Reliability-Guided Refinement — Medical image classification improves by integrating multimodal knowledge graphs from similar cases with reliability-guided refinement.
• Benchmarks and methods for 3D medical image retrieval — The 3D-MIR benchmark and methods advance medical image retrieval by providing standardized evaluation and multi-modal search strategies.
• Medical AI gets 66% worse when you use automated labels for training, and the benchmark hides it! [R][P] — A recent study on fairness in medical segmentation for breast cancer tumors revealed that AI models perform significantly worse for younger patients, a bias attributed to the qualitative nature of the…
• GloResNet: A lightweight 3D CNN with global topological features for preterm brain injury prediction — Lightweight 3D CNN with global topological features predicts preterm brain injury effectively.
• v227: Proceedings of Medical Imaging with Deep Learning 2023 — Volume 227 of "Medical Imaging with Deep Learning," held in Nashville, TN, USA, from July 10-12, 2023, presents significant advancements in applying deep learning across diverse medical imaging challenges.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.LG updates on arXiv.org.