Laboratory test results are central to detecting, monitoring, and managing health conditions, but conventional systems flag abnormalities by comparing single values against fixed reference ranges an approach that misses subtle deviations and overlooks relationships between multiple parameters. This work presents a machine learning system that learns normal patterns from historical laboratory data to detect anomalies. After preprocessing to handle missing values and inconsistencies, the system analyzes multiple parameters jointly, capturing correlations between tests rather than evaluating each in isolation, and tracks a patient's values over time to surface unusual trends. Each record receives an anomaly score, which stratifies results into normal, attention, and high-risk categories. An explain ability component identifies which parameters drove a given anomaly score, and the system flags results that appear internally inconsistent and may warrant verification. Multiple algorithms are compared using precision, recall, F1-score, and false-positive rate. This framework offers a more comprehensive alternative to threshold-based analysis, helping clinicians prioritize significant patterns for review. It is designed as a decision-support tool, not a replacement for medical diagnosis or clinical judgment. Planned extensions include personalized patient baselines, real-time monitoring, and EHR integration.

Dharini S R, Suji C, "AI-Based Anomaly Detection in Laboratory Test Results" International Journal of Scientific Research and Engineering Development, V9(6): Page(187-190) November - December 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.