A Health Care Clinical Data Platform for Rapid Deployment of Artificial Intelligence and Machine Learning Algorithms for Cancer Care and Oncology Clinical Trials

Aug 5, 2026

Upload by Liam

The paper presents the xCures platform, designed to aggregate, organize, structure, and normalize clinical EMR data across multiple care sites for near real-time access. It standardizes information to support clinical decision support, accelerate oncology research, and enable AI/ML algorithm development, with emphasis on precision oncology workflows. Patients and providers access outputs through web portals, while the system operates in HIPAA-compliant cloud infrastructure using NER, NLP, and LLM-based structuring and cataloging. The work also describes data governance via IRB-approved registry use.

LanguageEnglish

CategoryTechnology

FormatPDF

Pages4

Reading Time10 min

AccessFree to download

Introduction

Challenges for AI/ML in healthcare

Methods

xCures平台用于病历汇聚与数据结构化

平台数据流与标准化能力

数据治理与试验支持

INVITED COMMENTARY A Health Care Clinical Data Platform for Rapid Deployment of Artificial Intelligence and Machine Learning Algorithms for Cancer Care and Oncology Clinical Trials Soma Sengupta, Rohan Rao, Zachary Kaufman, Timothy J. Stuhlmiller, Kenny K. Wong, Santosh Kesari, Mark A. Shapiro, Glenn A. Kramer The xCures platform aggregates, organizes, structures, and normalizes clinical EMR data across care sites, utilizing advanced technologies for near real-time access. The platform generates data in a format to support clinical care, accelerate research, and promote artificial intelligence/ machine learning algorithm development, highlighted by a clinical decision support algorithm for precision oncology. Introduction Artificial intelligence (AI), which includes natural lan guage processing (NLP) and large language models (LLM) as well as machine learning (ML), is increasingly being applied to electronic medical records (EMR) to enhance health care delivery and research. These technologies facilitate the querying, understanding, and extraction of information from EMRs, which is often unstructured and voluminous. AI-based methods, including ML classifiers that automatically categorize data, have been effectively applied to query EMR, demonstrating high diagnostic accuracy across multiple organ systems [1] . NLP, particularly when combined with ML and deep learning, can be used to automatically analyze medical documents, extracting information and classifying it into predefined categories [2] . This can improve health care delivery by aiding clinicians in finding relevant information or gaining insights from medical reports and clinical research forms. Named entity recognition (NER) models, such as ClinSpacy and MedSpacy, have been developed for clinical NLP to identify and extract specific medical concepts such as drug names and dosages from clinic notes, which is crucial for converting unstructured data into a structured format for downstream analysis [3] . Similarly, ML techniques have been applied to extract real-world data variables from unstructured EMR, such as information on cancer diagnoses and treatments, which can be used for evidence generation in oncology [4] . EMRs contain valuable medical information but are often difficult for computers to understand due to jargon and abbreviations. Since Google’s release of the Transformer architecture, LLMs have entered mainstream use [5] . Following the public release of the LLM known as GPT-3, many organizations have generated models based on similar architecture to create bespoke LLMs trained on healthcare data to extract and interpret EMR, enabling more personalized clinical recommendations [6–7] . To make AI/ML solutions useful in health care and fulfill their promise of improving patient care, easing clinician workloads, and accelerating clinical research, several challenges remain [8] . First, there is a shortage of available EMR training data, due to data privacy and security concerns. Second, there are complex issues with training models to understand the mix of structured and unstructured data in the records. Third, there is great diversity in the data models underlying EMR systems. Finally, the cost of training a custom LLM means that hospitals and health systems should generally not attempt to create their own. Thus, a general purpose clinical LLM that could integrate with all or most existing health care IT is yet to be realized. Here, we describe a decentralized precision oncology platform for gathering, organizing, structuring, and standardizing longitudinal clinical data directly from EMRs for use in downstream AI and ML applications. Methods xCures Platform for Medical Record Aggregation and Data Structuring The xCures platform (summarized in Figure 1) providesan infrastructure that allows for near real-time clinical data Electronically published July 10, 2024. Address correspondence to Soma Sengupta, Bioinformatics 5115, UNC S

FAQs

xCures平台如何实现临床EMR数据的近实时聚合与结构化？

平台通过HIE在近实时聚合来自多来源的EMR，并使用NER、NLP和LLM模型对数据进行分类与结构化，形成标准化数据湖与元数据目录。

平台的数据如何支持癌症精准肿瘤学与临床决策？

平台将结构化数据用于临床决策支持，例如第二意见与分子肿瘤委员会，并将数据用于后续AI/ML应用的算法开发。

xCures平台如何在隐私与合规方面降低机构端负担？

平台完全运行在HIPAA合规的云基础设施中，从而缓解机构本地服务器的运维负担，并通过IRB批准的注册体系（XCELSIOR）进行数据使用与工具开发。

Aug 5, 2026

Upload by Liam