A Masked Language Model for Multi-Source EHR Trajectories Contextual Representation Learning

Summary, in English

Using electronic health records data and machine learning to guide future decisions needs to address challenges, including 1) long/short-term dependencies and 2) interactions between diseases and interventions. Bidirectional transformers have effectively addressed the first challenge. Here we tackled the latter challenge by masking one source (e.g., ICD10 codes) and training the transformer to predict it using other sources (e.g., ATC codes).

• Författare: Ali Amirahmadi; Mattias Ohlsson; Kobra Etminani; Olle Melander; Jonas Björk
• Redaktör: Maria Hagglund; Madeleine Blusi; Stefano Bonacina; Lina Nilsson; Inge Cort Madsen; Sylvia Pelayo; Anne Moen; Arriel Benis; Lars Lindskold; Parisis Gallos
• Avdelning/ar: Centrum för miljö- och klimatvetenskap (CEC); LU profilområde: Naturlig och artificiell kognition; eSSENCE: The e-Science Collaboration; Avdelningen för arbets- och miljömedicin; EpiHealth: Epidemiology for Health; EXODIAB: Excellence of Diabetes Research in Sweden; MultiPark: Multidisciplinary research focused on Parkinson´s disease; Kirurgi och folkhälsa; EPI@LUND
• Publiceringsår: 2023
• Språk: Engelska
• Sidor: 609-610
• Publikation/Tidskrift/Serie: Studies in Health Technology and Informatics
• Volym: 302
• Länkar: http://dx.doi.org/10.3233/SHTI230217
• Dokumenttyp: Konferensbidrag
• Förlag: IOS Press
• Ämne: Computer Science
• Nyckelord: deep learning; disease prediction; electronic health records; Masked language model; patient trajectories; representation learning
• Conference name: 33rd Medical Informatics Europe Conference: Caring is Sharing - Exploiting the Value in Data for Health and Innovation, MIE2023
• Conference date: 2023-05-22 - 2023-05-25
• Conference place: Gothenburg, Sweden
• Status: Published
• Forskningsgrupp: Surgery and public health; EPI@LUND
• ISBN/ISSN/Övrigt: ISSN: 1879-8365; ISSN: 0926-9630; ISBN: 9781643683881