230 likes | 242 Views
Learn about the interplay of diagnostic decision support and machine learning in skin lesion and rash diagnosis. Explore the Medical Cockpit design and its comprehensive medical knowledge database. Discover how VisualDx supports professionals and patients with point-of-care differential diagnosis.
E N D
Autonomous Medicine| 8:30-9:30 Diagnostic Decision Support And AI Art Papier MD CEO VisualDx Associate Professor of Dermatology University of Rochester
LEARNING OBJECTIVES • Define diagnostic decision support • Describe the interplay of decision support and machine learning of skin lesions and rashes
Decision Support: + > Machine Learning Artificial Intelligence Rule Based Systems
Medical Cockpit Design: + > • Medical images of all forms and variations • Comprehensive medical knowledge database • Visual search process and interface • Merge terminology, images, database, machine learning and interface
2 matching findings: Arthralgia (Joint Pain, Articular Pain), Fever (Febrile) Searchable by condition, medication or unique patient factors delivering point-of-care differential diagnosis, testing and therapy. Medicon • Reduce Cognitive Burden: • Visualization of Complexity • Transforming the “diagnostic list” • by allowing for comparison of disease features
Findings entered: blanching patch, targetoid Findings entered: facial palsy, arthralgia Differential shown in photo view Differential shown in Sympticons Supports the recognition of the variation of disease presentation Example: Phases of Lyme Disease
VisualDx Multiple Workflows • Uses SMART on FHIR • Native iOS and Android apps • Interoperable in: • EHR • UpToDate • Telemedicine VisualDx iOS and Android Apps VisualDx within UpToDate VisualDx on modern browsers VisualDx is in Cerner
International network ofdedicated contributors has participated in building good data. Images and case data are submitted by experts from around the world. Diseases of regional and geographic importance are captured, uploaded and labeled
The Basis For Machine Learning and AI: Good Data Human Phenome From genome to transcriptome to proteome to microbiome, the importance of these efforts ultimately hinges on connecting the genetic and molecular data to the clinical manifestations of disease. Machine Learning The goal of our research is to merge image recognition with knowledge databases to improve diagnosis.
Machine Learning: Training on Skin Examination Description
Recent Case Study • Presents to ED on April 10, 2018 • Female, 66 years old • Two-week history of enlarging lesion on finger. Patient removed lesion herself. Lesion recurred larger. Brian Browne, MD
Machine Learning Review diagnostic possibilities. Snap a picture. Confirm lesion type. Add other symptoms. Process
Apple’s Core ML: Speed and on Device
VisualDx is Widely Used Point of Care Tool 59 Images viewed in 2017 MILLION+ 2,695,470 Clinical Inquiries
What about AI for Non-Physicians?
Aysa • Aysa gives people personalized guidance about what to do for a set of 200 common skin conditions. • Aysa is focused on skin conditions because this is an area of real need and a place where we have the expertise to help. See results. Answer questions. Review information. Take a photo.