NeuroScan AI: An Explainable Deep Learning Framework for Brain Hemorrhage Detection and Clinical Decision Support Using CT Scan Images

Abstract

Brain hemorrhage is a life-threatening neurological emergency in
which rapid diagnosis directly affects patient survival and long-term
recovery. In current clinical practice, CT scans are read manually by
radiologists, a process that is accurate but slow and heavily
dependent on the availability of experienced specialists, particularly
in busy emergency departments. This paper introduces NeuroScan
AI, an explainable deep learning framework built to detect brain
hemorrhage from CT scans and support clinicians with automated,
easy-to-understand diagnostic reasoning. At its core, the system uses
a Convolutional Neural Network (CNN) to classify CT images as
hemorrhage or non-hemorrhage. What sets NeuroScan AI apart is
that its predictions are not left as raw labels — they are passed
through OpenAI's language model to produce clear, clinically
relevant explanations that help physicians understand why a given
classification was made. The framework is delivered as a secure,
web-based platform that also stores patient records and diagnostic
reports for later reference. In testing, the system produced fast,
accurate, and interpretable predictions, shortening the time needed
to reach a preliminary diagnosis while giving clinicians more
confidence in the AI's output. These results suggest that NeuroScan
AI could meaningfully improve the efficiency, reliability, and
accessibility of brain hemorrhage diagnosis in real-world healthcare
settings.