Data Curation for AI with Proper Medical Imaging Physics Context (2022)-M3-CHP09-2022

Video Transcription

Video Summary

This video explores the crucial role of high-quality data in developing AI models for clinical use, emphasizing its impact on patient care quality and safety. Starting with a discussion on preparing high-quality data, the talk covers the steps of planning, acquiring, and utilizing data, particularly in the context of chest X-ray-based COVID-19 diagnosis using AI. It stresses the importance of collaboration, data sharing, and using data tools such as RSNA-CTP for effective data management.<br /><br />The second part highlights why high-quality data is pivotal for AI and its generalizability. It covers data quality assurance considerations, including maintaining native DICOM format, accurate labeling using time-window restricted RT-PCR results, and ensuring equal sourcing of COVID-positive and negative data.<br /><br />Finally, the discussion shifts to AI model deployment in clinical settings. It touches on technical aspects of deployment in PACS and EMR systems, considering factors like imaging data, EMR data integration, and orchestration engines. The importance of monitoring, defining performance metrics, ensuring fairness, and making continuous improvements to ensure AI models' clinical success are underscored, highlighting AI's potential to enhance imaging processes when carefully and thoughtfully integrated.

Keywords

high-quality data

AI models

clinical use

patient care

COVID-19 diagnosis

data management

DICOM format

AI deployment

EMR integration