The Large Cancer Assistant (LCA) Model-Agnostic Orchestration Framework for Scalable Clinical Decision Support in Oncology

Jul 16, 2026

Upload by Sage

The Large Cancer Assistant (LCA) is a model-agnostic, post-hoc orchestration framework for scalable clinical decision support in oncology, addressing limitations of monolithic multimodal systems that tightly couple data ingestion, clinical routing, and AI inference. LCA is formalized as a 7-tuple architecture under Algorithmic Impermeability, keeping orchestration logic independent from underlying black-box models. Entry Theory standardizes multimodal patient data, and a Cancer Switching Module produces a Standardized Intermediate Payload (SIP) to enable EMR isolation and interoperability.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages22

Reading Time55 min

Introduction

Multimodal nature of oncology

Limitations of monolithic multimodal systems

Informatics orchestration perspective

Highlights The Large Cancer Assistant (LCA): A Model-Agnostic Orchestration Framework for Scalable Clinical Decision Support in Oncology Ghassen MARRAKCHI, Basarab MATEI • LCA orchestrates multimodal oncology data without model coupling. • Algorithmic impermeability strictly isolates AI from data routing. • Entry Theory algebraically standardizes diverse clinical formats. • A Standardized Intermediate Payload (SIP) ensures EMR isolation. • Proof of concept proves 100% failure safety and zero AI overhead. arXiv :2607 .0653 1v 1 [ cs .AI ] 7 Jul 2026 The Large Cancer Assistant (LCA): A Model-Agnostic Orchestration Framework for Scalable Clinical Decision Support in Oncology Ghassen MARRAKCHIa,∗ , Basarab MATEIa aLIPN, CNRS, UMR 7030— Université Sorbonne Paris Nord, 99 Av. Jean Baptiste Clément, Villetaneuse, F-93430, France | ARTICLE INFO | | AB STRACT | | | --- | --- | --- | --- | | Keywords: clinical decision support multimodal orchestration algorithmic impermeability oncology informatics interoperability model-agnostic architecture | | Objective: Multimodal deep learning models in oncology are currently limited by monolithic designs that rigidly couple data ingestion, clinical routing, and artificial intelligence (AI) inference. To address this inflexibility, we propose the Large Cancer Assistant (LCA), a model-agnostic, post-hoc orchestration framework designed for scalable clinical decision support. Methods: The LCA is mathematically formalized as a 7-tuple architecture grounded in the principle of Algorithmic Impermeability, ensuring the orchestration logic remains strictly independent of underlying black-box AI models. We introduce the Entry Theory, leveraging Geometric Deep Learning (GDL) to standardize multimodal patient data along distinct structural and medical axes. The system dynamically orchestrates data via a Cancer Switching Module and intentionally isolates the core AI execution from volatile hospital IT infrastructures by outputting a Standardized Intermediate Payload (SIP) . Results: A Proof of Concept (PoC) validated the orchestration logic across four technical scenarios. The framework executed a nominal flow with negligible orchestration overhead. It empirically demonstrated algorithmic impermeability by maintaining an invariant routing projection during AI model swaps, and it validated strict failure-safety by achieving a 100% recall rate in generating targeted Supplementary Data Requests (SDR) under injected data anomalies. Multi-protocol execution capability was also successfully verified. Conclusion: By structurally decoupling multimodal ingestion from feature inference, the LCA provides a highly adaptable and modular orchestration foundation. The SIP establishes a clear architectural boundary, natively setting the stage for downstream Electronic Medical Record (EMR) interoperability as an independent future paradigm. | | | 1. Introduction The landscape of clinical oncology is inherently multimodal. A comprehensive patient diagnosis rarely relies on a single data source; rather, it requires the continuous synthesis of high-dimensional spatial imaging (e.g., computed tomography, magnetic resonance imaging), unstructured semantic histories (e.g., clinical notes, pathology reports), and tabular biological metrics (e.g., blood panels) . Consequently, the integration of Artificial Intelligence (AI) into oncology has largely focused on developing multimodal deep learning models capable of processing these diverse data streams to improve diagnostic and prognostic accuracy. However, the clinical translation of these advanced models is currently constrained by systemic architectural flaws. Many existing multimodal systems remain tightly coupled to specific modality combinations and task definitions, limiting their flexibility across heterogeneous clinical workflows. They tightly couple the ingestion of data, the routing logic, and the core neural network inference

FAQs

What problem does the Large Cancer Assistant (LCA) address in oncology clinical decision support?

LCA targets the inflexibility of existing multimodal systems that rigidly couple data ingestion, routing, and AI inference into a single monolithic pipeline, limiting adaptability across heterogeneous clinical workflows.

How does LCA remain model-agnostic and prevent interference with black-box AI models?

LCA formalizes orchestration using the principle of Algorithmic Impermeability, ensuring the orchestration logic stays strictly independent of the underlying AI inference models.

What role do Entry Theory and SIP play in LCA’s architecture?

Entry Theory algebraically standardizes diverse clinical formats, and the Standardized Intermediate Payload (SIP) isolates EMR interactions so that downstream interoperability can be handled through a clear architectural boundary.

Jul 16, 2026

Upload by Sage