Näytä suppeat kuvailutiedot

Adaptive risk prediction system with incremental and transfer learning

Sahota Daljit Singh; Airola Antti; Lo Tsz-kin; Leung Wing-cheong; Koivu Aki; Sairanen Mikko; Pahikkala Tapio

dc.contributor.authorSahota Daljit Singh
dc.contributor.authorAirola Antti
dc.contributor.authorLo Tsz-kin
dc.contributor.authorLeung Wing-cheong
dc.contributor.authorKoivu Aki
dc.contributor.authorSairanen Mikko
dc.contributor.authorPahikkala Tapio
dc.date.accessioned2022-10-28T14:27:44Z
dc.date.available2022-10-28T14:27:44Z
dc.identifier.urihttps://www.utupub.fi/handle/10024/171501
dc.description.abstract<p><br></p><p>Currently, popular methods for prenatal risk assessment of fetal aneuploidies are based on multivariate proba-bilistic modelling, that are built on decades of scientific research and large-scale multi-center clinical studies. These static models that are deployed to screening labs are rarely updated or adapted to local population characteristics. In this article, we propose an adaptive risk prediction system or ARPS, which considers these changing characteristics and automatically deploys updated risk models. <br></p><p>8 years of real-life Down syndrome screening data was used to firstly develop a distribution shift detection method that captures significant changes in the patient population and secondly a probabilistic risk modelling system that adapts to new data when these changes are detected. Various candidate systems that utilize transfer-and incremental learning that implement different levels of plasticity were tested. <br></p><p>Distribution shift detection using a windowed approach provides a computationally less expensive alternative to fitting models at every data block step while not sacrificing performance. This was possible when utilizing transfer learning. Deploying an ARPS to a lab requires careful consideration of the parameters regarding the distribution shift detection and model updating, as they are affected by lab throughput and the incidence of the screened rare disorder. When this is done, ARPS could be also utilized for other population screening problems. <br></p><p>We demonstrate with a large real-life dataset that our best performing novel Incremental-Learning-Population-to-Population-Transfer-Learning design can achieve on par prediction performance without human intervention, when compared to a deployed risk screening algorithm that has been manually updated over several years.</p>
dc.language.isoen
dc.publisherPERGAMON-ELSEVIER SCIENCE LTD
dc.titleAdaptive risk prediction system with incremental and transfer learning
dc.identifier.urnURN:NBN:fi-fe2021102952989
dc.relation.volume138
dc.contributor.organizationfi=terveysteknologia|en=Terveysteknologia|
dc.contributor.organizationfi=tietotekniikan laitoksen yhteiset|en=Tietotekniikan laitoksen yhteiset|
dc.contributor.organizationfi=data-analytiikka|en=Data-analytiikka|
dc.contributor.organization-code2610300
dc.contributor.organization-code2610301
dc.contributor.organization-code2610303
dc.converis.publication-id67532320
dc.converis.urlhttps://research.utu.fi/converis/portal/Publication/67532320
dc.identifier.jour-issn0010-4825
dc.okm.affiliatedauthorKoivu, Aki
dc.okm.affiliatedauthorAirola, Antti
dc.okm.affiliatedauthorPahikkala, Tapio
dc.okm.discipline113 Tietojenkäsittely ja informaatiotieteetfi_FI
dc.okm.discipline113 Computer and information sciencesen_GB
dc.okm.internationalcopublicationinternational co-publication
dc.okm.internationalityInternational publication
dc.okm.typeJournal article
dc.publisher.countryBritanniafi_FI
dc.publisher.countryUnited Kingdomen_GB
dc.publisher.country-codeGB
dc.relation.articlenumberARTN 104886
dc.relation.doi10.1016/j.compbiomed.2021.104886
dc.relation.ispartofjournalComputers in Biology and Medicine
dc.year.issued2021


Aineistoon kuuluvat tiedostot

Thumbnail

Aineisto kuuluu seuraaviin kokoelmiin

Näytä suppeat kuvailutiedot