Lightweight model achieves broad medical image segmentation with 211k parameters
A 211K Parameter Foundation Model: Genetic Prompting in Neural Cellular Automata
Computer Vision and Pattern Recognition
Summary
Medical image segmentation is important for diagnosing many illnesses, but typically requires large computer models that need powerful equipment. The authors created a very small model called Foundational-NCA that can handle many different segmentation tasks without retraining. It uses a special technique combining cellular automata and genetic prompting to adapt dynamically to various tasks. This small model performs well on both known and new tasks, making it easier to use in places with limited computing resources.
What this means in practice
- •For medical device developers: Develop compact AI tools for medical image segmentation that operate effectively in hospitals with limited computing infrastructure.$Commercial implications: Enables creation of affordable medical imaging products for low-resource healthcare settings by reducing AI model complexity dramatically.
- •For mobile health app developers: Integrate lightweight segmentation models into mobile apps for diagnostic support without demanding high-end hardware.
Authors
John Kalkhof, Niklas Ihm, Daniel Kirste, Anirban Mukhopadhyay
Abstract
The scale of modern foundation models enables broad task generalization. In medical imaging, where segmentation demands vary widely across clinical contexts, such flexibility is invaluable. But current models, such as SAM trade adaptability for scale, locking their utility behind high-end infrastructure and leaving behind the low-resource environments most in need of diagnostic support. To bridge this gap, we introduce Foundational-NCA, a lightweight foundation model capable of adapting to a wide range of medical segmentation tasks without retraining. Built on Neural Cellular Automata (NCA), Foundational-NCA incorporates gradient-based learning with genetic optimization over a latent task space, using a compact task vector to guide predictions. This enables dynamic adaptation via a genetic prompt mechanism, allowing the model to generalize across tasks while keeping the parameter count magnitudes lower at just 211 thousand, compared to SAM-based models with 38.9 to 636 million parameters. We evaluate Foundational-NCA on 15 trained and 6 entirely unseen segmentation tasks from the TotalSegmentator dataset, achieving an average Dice score of 84.5% on known tasks and 60.4% on unseen ones. These results demonstrate that task generalization is possible in ultra-compact models, laying the groundwork for resource-efficient, deployable foundation models in clinical environments where infrastructure is limited. By reducing model size by over 180 times, Foundational-NCA takes a step toward a more equitable future for medical AI, one where powerful tools are within reach of every healthcare setting.