Evaluating the generalisability of region-naïve machine learning algorithms for the identification of epilepsy in low-resource settings
Clinical models for epilepsy diagnosis in LMICs demonstrate characteristic traits of ML models, such as limited generalisability and a trade-off between internal and external performance. The relationship between predictors and model outcomes also varies across sites, suggesting the need to update specific model aspects with local data before broader implementation. Variations are likely to be particular to the cultural context of diagnosis. We recommend developing models adapted to the cultures and contexts of their intended deployment and caution against deploying region- and culture-naïve models without thorough prior evaluation.
Authors
- Judith Helen Cross (ORCID: https://orcid.org/0000-0001-7345-4829)
- Gabriel Davis Jones (ORCID: https://orcid.org/0000-0003-4070-4814)
- Seth Owusu‐Agyei (ORCID: https://orcid.org/0000-0002-0807-1781)
- Daniel Mwanga (ORCID: https://orcid.org/0000-0003-3907-1538)
- Charles R. Newton (ORCID: https://orcid.org/0000-0002-6999-5507)
- Honorati Masanja (ORCID: https://orcid.org/0000-0002-3860-7348)
- Anthony Ngugi (ORCID: https://orcid.org/0000-0001-5415-8717)
- Angelina Kakooza‐Mwesige (ORCID: https://orcid.org/0000-0002-1189-3671)
- Josemir W. A. S. Sander (ORCID: https://orcid.org/0000-0001-6041-9661)
- Arjune Sen (ORCID: https://orcid.org/0000-0002-8948-4763)
- Ryan G. Wagner (ORCID: https://orcid.org/0000-0003-2741-3676)
- Symon M. Kariuki (ORCID: https://orcid.org/0000-0002-3934-5132)
- Ioana Duta (ORCID: https://orcid.org/0009-0000-1278-8673)
Institutions
- Ifakara Health Institute (TZ)
- University of Nairobi (KE)
- University of the Witwatersrand (ZA)
- Sichuan University (CN)
- University of Health and Allied Sciences (GH)
- Pwani University (KE)
- John Radcliffe Hospital (GB)
- Kenya Medical Research Institute (KE)
- University of Oxford (GB)
- West China Hospital of Sichuan University (CN)
- Agincourt Health and Socio-Demographic Surveillance System (ZA)
- Aga Khan University Nairobi (KE)
- Indepth Network (GH)
- The Alan Turing Institute (GB)
- Kintampo Health Research Centre (GH)
- Stichting Epilepsie Instellingen Nederland (NL)
- University College London (GB)
- Epilepsy Society (GB)
- Makerere University (UG)
- Wellcome Trust (GB)
Publication Details
- Journal
- PLOS Digital Health
- Published
- 2025-02-12
- DOI
- https://doi.org/10.1371/journal.pdig.0000491
- Citations
- 4
- Primary Topic
- Epilepsy research and treatment
- Type
- article
- Field-Weighted Citation Impact
- 5.88
Funders
- National Institute for Health and Care Research
- University of Oxford