Machine Learning & Medical AI Researcher · Deep Learning for Medical Imaging
Building deep-learning systems that learn from limited data and make robust, accurate decisions — with applications across medical imaging and education.
Selorm Adablanu is a computer scientist and machine-learning researcher whose work centres on deep learning for medical imaging and the responsible application of artificial intelligence in healthcare and education. His recent research develops hybrid transformer–CNN architectures for stroke and skin-cancer detection, and rigorously evaluates how optimizers, loss functions, and class-imbalance strategies shape model reliability in clinical settings. Alongside primary modelling work, he has authored PRISMA-guided systematic reviews that map the methodological landscape of medical deep learning. A consistent thread runs through his contributions: designing AI systems that learn effectively from limited data and remain robust, interpretable, and reproducible. He is currently pursuing a Ph.D. in Computer Science and Engineering.
A deep-learning architecture combining Vision Transformers, CNNs and LSTM modules with a meta-learning component to capture spatial, temporal and global features from CT stroke images; later adapted to dermoscopic skin-cancer classification.
A PRISMA-guided systematic review of 69 studies (2010–2025) analysing how adaptive, momentum-based, and metaheuristic optimizers affect classification, segmentation and detection across medical imaging modalities.
A comparative study of Focal Loss versus Binary Cross-Entropy for class-imbalanced stroke CT classification, examining sensitivity, AUC and the effects of test-time augmentation on an Xception backbone.
Interdisciplinary work on artificial intelligence, mobile and IoT-enabled learning, and multimedia-supported instruction in higher education, drawing on Ghanaian institutional case studies.
Neuroscience Informatics
First systematic review of optimization algorithms across 69 medical deep-learning studies, with task- and architecture-specific recommendations for reproducible deployment.
iRADIOLOGY
Adapts the hybrid VINCE-NET model to melanoma classification on the HAM10000 dataset, achieving competitive accuracy while remaining feasible in resource-limited settings.
6th Int. Conference on Advances in Electrical, Computing, Communications & Sustainable Technologies (ICAECT 2026)
Shows Focal Loss improves sensitivity on imbalanced stroke CT data and cautions that test-time augmentation must match training-time augmentation.
Feature Selection and Feature Extraction on Omics Data (book chapter)
Examines how dimensionality-reduction and feature-engineering strategies compare with end-to-end models on high-dimensional omics data.
Discover Neuroscience
Introduces VINCE-NETv1, validated across multiple CT stroke datasets with patient-level separation and early Grad-CAM interpretability evidence.
Brain Hemorrhages
Critically reviews 34 studies (2014–2025) spanning traditional ML, deep, transfer and hybrid learning for stroke detection, with directions for explainable, clinically deployable AI.
This list syncs automatically from the open ORCID registry (with live citation counts from Crossref) each time the page loads; the six Scopus-indexed records above are the verified baseline shown whenever the live feed is unavailable. A fuller list is on Google Scholar and ResearchGate. Citation counts are point-in-time.
Verified reviewer record (MVW-3570-2025). Invited reviewer for the following indexed journals:
Editorial contributor — Journal of King Saud University – Computer and Information Sciences (Elsevier).
Build Back Better COMPASS Workshop 2023 — “Approaches to Leveraging Digital Higher Education in Africa.” · Advanced Excel Using AI Tools — Nerim Group of Institutes, Guwahati, India.
Open to research collaboration, academic peer review, postgraduate research opportunities, and interdisciplinary projects in AI, healthcare, and data-driven research — including medical imaging, biomedical machine learning, systematic reviews and evidence synthesis, bibliometric studies, and AI in education.