CARAT (Computer-Aided Radiology AI Tool) is a brain tumor MRI classification model trained on 5,712 MRI scans across 4 classes: Glioma, Meningioma, Pituitary, and No Tumor. Built with a CNN and deployed as a TFLite model on a Raspberry Pi 5, it gi…
CARAT (Computer-Aided Radiology AI Tool) is a brain tumor MRI classification model trained on 5,712 MRI scans across 4 classes: Glioma, Meningioma, Pituitary, and No Tumor. Built with a CNN and deployed as a TFLite model on a Raspberry Pi 5, it gives clinicians in rural or low-resource settings instant AI-assisted diagnoses without any internet connection. Part of the radpi medical diagnostic platform.
Used Perplexity AI for research and architecture decisions. GitHub Copilot assisted with model training code and TFLite conversion scripts.