The Large Cancer Assistant (LCA) is a model-agnostic, post-hoc orchestration framework designed to address the inflexibility of monolithic multimodal deep learning models in oncology. It decouples data ingestion from clinical routing and AI inference using a 7-tuple architecture grounded in Algorithmic Impermeability.

  • The system leverages Geometric Deep Learning via Entry Theory to standardize multimodal patient data along distinct structural and medical axes.
  • A Cancer Switching Module dynamically orchestrates data while isolating core AI execution from volatile hospital IT infrastructures.
  • Proof of Concept validation demonstrated negligible orchestration overhead and maintained invariant routing projections during AI model swaps.
  • The framework achieved a 100% recall rate in generating targeted Supplementary Data Requests under injected data anomalies.

By establishing a Standardized Intermediate Payload, the LCA provides a modular foundation for scalable clinical decision support and sets the stage for downstream Electronic Medical Record interoperability.