Low Dose CT for Stroke Diagnosis: A Dual Pipeline Deep Learning Framework for Portable Neuroimaging

TL;DR

This paper introduces a deep learning framework for stroke diagnosis using low-dose portable CT scans, comparing direct classification and denoising pipelines, and evaluates their performance across different dose levels.

Contribution

It presents a dual pipeline approach for LDCT stroke classification and provides a reproducible baseline for AI-assisted triage in portable neuroimaging.

Findings

Denoising improves perceptual image quality but not always classification accuracy.

Direct classification often yields higher sensitivity than denoising-then-classify.

Best pipeline achieves 0.94 AUC and 0.91 accuracy at moderate dose levels.

Abstract

Portable CT scanners enable early stroke detection in prehospital and low-resource settings but require reduced radiation doses, introducing noise that degrades diagnostic reliability. We present a deep learning framework for stroke classification from simulated low-dose CT (LDCT) brain scans for AI-assisted triage in mobile clinical environments. Controlled Poisson noise is applied to high-dose CT images to simulate realistic LDCT conditions. We compare two pipelines: (1) direct classification of noisy LDCT images and (2) denoising followed by classification. Performance is evaluated across multiple dose levels using accuracy, sensitivity, and AUC. While denoising improves perceptual image quality, it does not consistently improve classification. In several settings, direct classification yields higher sensitivity, revealing a trade-off between perceptual quality and diagnostic utility. The best denoise-then-classify pipeline achieves 0.94 AUC and 0.91 accuracy at moderate dose levels, outperforming direct classification by up to 6% in select cases. This work establishes a reproducible baseline for LDCT stroke triage using hemorrhagic stroke data (RSNA dataset) and highlights the need for validation on ischemic cohorts and real-world portable CT systems.