A Hydrogen-Supported Framework for Detecting Lacunar Infarcts Using Neuroimaging and Deep-Learning Algorithms
Lacunar infarction is a form of cerebral small-vessel disease (SVD) and is often not identified, as lesions can be small, subtle and morphologically indistinguishable from other white-matter processes. This manuscript introduces a framework for detecting lacunar infarcts in neuroimaging data based on medical-image preprocessing, conventional machine learning, convolutional neural networks, and residual neural networks, all with the assistance of hydrogen. There are two interrelated, but analytically distinct branches of the framework. The diagnostic branch collects and processes anonymized MRI or CT data, builds predictive models, and assesses the models using appropriate metrics such as sensitivity, specificity, precision, recall, F1 score, area under the receiver-operating-characteristic curve, calibration and lesion-localisation metrics. The infrastructure branch estimates the electricity consumption for image acquisition, data storage, model training, cooling, and clinical inference, and evaluates the technical and environmental feasibility of using hydrogen as a back-up or long-duration storage method. Hydrogen is not suggested as a substitute for clinical skills or as a treatment or diagnostic signal. Its intended use is for reliable and resilient operation of infrastructure needed for the deployment of medical-AI. This is a proposed framework and not a completed clinical validation study, therefore, the manuscript does not make any claims of the effectiveness of the diagnosis. Instead, it outlines a reproducible design and an evaluation program to use in future empirical research.
Authors
- Bukola Badeji–Ajisafe (ORCID: https://orcid.org/0000-0002-0944-4615)
- Kehinde Abiola
- Jumoke, OLUSOLA, Duyilemi
- Abimbola Omotola Ajibade
- Stephen Eyitayo Obamiyi
Institutions
- Afe Babalola University (NG)
- Federal University of Technology and Environmental Sciences (NG)
Publication Details
- Journal
- International Journal of Energy Research Science and Engineering
- Published
- 2026-11-17
- DOI
- https://doi.org/10.53982/ijerse.2026.0102.06-j
- Primary Topic
- Artificial Intelligence in Healthcare and Education
- Type
- article
- Field-Weighted Citation Impact
- 0.00