A context-aware mobile platform for drug interaction alerts and patient-centered insight

Tan, Shin Yi (2026) A context-aware mobile platform for drug interaction alerts and patient-centered insight. Final Year Project, UTAR.

PDF Download (6Mb)

Abstract

Medication safety remains a critical healthcare challenge, including risks such as pill misidentification, drug interactions, allergic reactions, scheduling conflicts, and missed-dose errors. This project develops a unified mobile health application integrating six intelligent modules to address these issues. The system includes pill recognition using ResNet-18 and ResNet-50, drug–drug interaction detection with AI explanations, allergy checking via ingredient matching, real-time drug information retrieval from OpenFDA and DrugBank, pharmacokinetic-aware medication scheduling, and a T-learner-based missed dose decision model trained on 10,000 synthetic patients. Results show 91.2% accuracy for pill recognition, 0% interaction risk under valid schedules, and a missed-dose model performance of PEHE 7.18, 63.7% accuracy, and 5.2% improvement over rule-based methods. The novelty lies in integrating six safety modules into a single platform, applying causal learning for missed-dose decisions, and combining pharmacokinetic-aware scheduling with dual CNN-based pill recognition. The system enhances medication safety through an integrated, user-centered framework.

Actions (login required)