Health-Guard AI: Integrated Health Report Management and Analysis

Abstract

This paper introduces Health-Guard AI, the current standard web-based application of health report management and analysis. Made up with Flask, MySQL and machine learning, the system allows users to upload files in the form of PDF health reports, automatically extract and push test data into a personalized interactive dashboard and attain recommendations by AI. The data used were the 500 health reports that had been II anonymized into Complete Blood Count (CBC) tests, results of lipid profile, and collected in collaboration with a partner diagnostic facility. The data in Reports were processed with the help of PDF-to-text conversion and extraction of tabular data and normalized according to references ranges, which presupposed privacy, as confidential information was discarded. A Logistic Regression model was then trained on this data to predict the test results as Low, Normal or High. A high accuracy of 92.4% and precision of 91.8%, recall of 93.1 and F1-score of 92.4 were achieved using as-fold cross- validation method. These findings illustrate that Health-Guard AI can help to streamline health data management processes and minimize the number of human errors that take place in the process of medical decision- making.

Ähnliche Arbeiten