Facilitating Healthcare Access and Services through AI-driven Medical Analysis
Bibliographic record
Abstract
A significant portion of the global population lacks access to essential healthcare services, particularly in remote and underserved regions, where traditional healthcare systems are often constrained by limited infrastructure and high patient costs.This thesis investigates how Artificial Intelligence (AI) can address these disparities and support the United Nations' Sustainable Development Goal 3 (SDG3) for universal health coverage by enabling intelligent, scalable, and accessible medical services.Two innovative methods are explored: In-Context Evolutionary Search (ICE-SEARCH), an LLM-based framework that integrates evolutionary algorithms with in-context learning to enhance medical feature selection and predictive analytics; and AI Clinics on Mobile (AICOM), a lightweight deployment solution that delivers monkeypox diagnostic AI tools on low-resource mobile devices without internet access.ICE-SEARCH significantly improves predictive accuracy for medical conditions such as stroke, cardiovascular disease, and diabetes by leveraging LLMs to identify critical features from complex datasets.Meanwhile, AICOM ensures healthcare delivery in areas with limited connectivity by reducing model size and computational demands while preserving user Abstract ii privacy.Although key challenges remain, this work demonstrates the potential of medical AI to democratize healthcare and enhance advanced medical analysis and service delivery across clinical domains.Future research directions are outlined to guide ongoing progress in this evolving field.iii Abrégé Une part importante de la population mondiale n'a pas accès aux services de santé essentiels, en particulier dans les régions éloignées et mal desservies, où les systèmes de santé traditionnels sont souvent limités par une infrastructure insuffisante et des coûts élevés pour les patients.Ce mémoire examine comment l'intelligence artificielle (IA) peut contribuer à réduire ces inégalités et soutenir l'Objectif de Développement Durable 3 des Nations Unies, visant une couverture santé universelle, en rendant les services médicaux plus intelligents, évolutifs et accessibles.Deux méthodes innovantes sont explorées : In-Context Evolutionary Search (ICE-SEARCH), un cadre basé sur les modèles de langage de grande taille (LLM) qui intègre des algorithmes évolutionnaires avec l'apprentissage contextuel pour améliorer la sélection de caractéristiques médicales et les analyses prédictives ; et AI Clinics on Mobile (AICOM), une solution légère de déploiement qui permet d'utiliser des outils de diagnostic de la variole simienne sur des appareils mobiles à faibles ressources, sans connexion Internet.ICE-SEARCH améliore considérablement la précision prédictive pour des pathologies telles que les AVC, les maladies cardiovasculaires et le diabète, en exploitant les LLM pour identifier Abstract iv les caractéristiques cliniques critiques à partir de jeux de données complexes.De son côté, AICOM permet l'accès aux soins dans les zones à connectivité limitée, en réduisant la taille des modèles et les besoins en calcul, tout en garantissant la confidentialité des utilisateurs.Bien que des défis importants subsistent, ce travail démontre le potentiel de l'IA médicale à démocratiser l'accès aux soins et à améliorer l'analyse médicale avancée et la prestation de services dans les domaines cliniques.Des pistes de recherche futures sont proposées afin d'orienter les progrès continus dans ce domaine en pleine évolution.v Large language model tools, such as ChatGPT, were used to improve the clarity, grammar, and overall phrasing of this thesis.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.011 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.004 | 0.003 |
| Open science | 0.002 | 0.004 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".