Leveraging large language models for enhanced diagnosis of neuroendocrine tumours

#4522

Introduction: Neuroendocrine tumours (NETs) are a heterogeneous group of cancers that are difficult to diagnose due to their variable clinical presentation and non-specific symptoms. Accurate diagnosis often requires integrating a wide range of clinical, radiological, and pathological information. However, clinicians may struggle to synthesise these diverse data sources effectively. Large language models (LLMs) have shown promise in processing and understanding complex medical text, making them a valuable tool for improving diagnostic accuracy in oncology.

Aim(s): This study explores the potential of LLMs to assist in the diagnosis of neuroendocrine tumours by integrating and analysing clinical notes, radiological reports, and pathology results to provide a more comprehensive and accurate assessment.

Materials and methods: We utilised a dataset consisting of patient records, including clinical documentation, imaging interpretations, and pathological findings. An LLM was fine-tuned on this data to learn how to extract and synthesise relevant information from various sources. The model was trained to identify patterns and relationships that could assist in diagnosing NETs, using advanced natural language processing (NLP) techniques to interpret medical terminology and clinical context.

Conference:

Presenting Author:

Authors: Tang Z, Tang J, Cheng F, Li H, Chen Y,

Keywords: Neuroendocrine Tumour (NET), Large Language Model (LLM), Medical Natural Language Processing (NLP), Clinical Decision Support,