Can Coding Agents Build Robust Baselines? A Skill-Based Approach for Automating the Medical Imaging Model-Development Pipeline
2026-08-24 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors created an AI system that helps build strong baseline models for medical imaging tasks by reading research papers, writing code, and running experiments automatically. Their system works across different medical imaging problems like segmentation, classification, and detection without needing special adjustments for each task. When tested on several public challenges, the AI-generated models performed well and showed good adaptability to different conditions. This approach reduces the manual work usually needed to develop competitive models in medical imaging.
deep learningmedical imagingbaseline modelssegmentationclassificationdetectionautomated machine learninghyperparameter tuningdomain generalizationAI Scientist
Authors
Eugenia Moris, José Ignacio Orlando
Abstract
Developing competitive deep learning baselines for medical imaging remains a highly iterative process requiring literature review, implementation, experimentation, and expert refinement. Existing automation approaches typically optimize isolated components, such as architecture search or hyperparameter tuning, rather than the complete baseline development process. We present an agentic AI Scientist workflow that combines literature-guided reasoning, automated code generation, and hypothesis-driven experimentation to generate competitive baseline models for medical imaging challenges. The framework is evaluated on four public benchmarks spanning segmentation, classification, and detection. Across all tasks, the Experimentation Pipeline consistently improves validation performance, achieving competitive leaderboard results, including 6th place on both PUMA tracks (15 teams) and 31st place on MILK10k (125 teams). On MIDOG25, the resulting model also demonstrates strong domain generalization across scanners, tumor types, and species. Using the same workflow across all challenges without task-specific redesign, we demonstrate that skill-based, literature-guided agentic workflows can substantially reduce the engineering effort required to develop competitive medical imaging baselines.