Abstract
PurposeAdvanced machine-learning (ML) techniques can potentially detect the entire spectrum of pathology through deviations from a learned norm.We investigated the utility of a weakly supervised ML tool to detect characteristic findings related to ischemic stroke in head CT and provide subsequent patient triage.MethodsPatients having undergone non-enhanced head CT at a tertiary care hospital in April 2020 with either no anomalies, subacute or chronic ischemia, lacunar infarcts of the deep white matter or hyperdense vessel signs were retrospectively analyzed. Anomaly detection was performed using a weakly supervised ML classifier. Findings were displayed on a voxel-level (heatmap) and pooled to an anomaly score. Thresholds for this score classified patients into i) normal, ii) inconclusive, iii) pathological. Expert-validated radiological reports were considered as ground truth. Test assessment was performed with ROC analysis; inconclusive results were pooled to pathological predictions for accuracy measurements.ResultsDuring the investigation period 208 patients were referred for head CT of which 111 could be included. Definite ratings into normal/pathological were feasible in 77 (69.4%) patients. Based on anomaly scores, the AUC to differentiate normal from pathological scans was 0.98 (95% CI 0.97–1.00). The sensitivity, specificity, positive and negative predictive values were 100%, 40.6%, 80.6% and 100%, respectively.ConclusionOur study demonstrates the potential of a weakly supervised anomaly-detection tool to detect stroke findings in head CT. Definite classification into normal/pathological was made with high accuracy in > 2/3 of patients. Anomaly heatmaps further provide guidance towards pathologies, also in cases with inconclusive ratings.
Highlights
Computed tomography (CT) of the head remains the primary modality in stroke imaging
Beyond the mere exclusion of hemorrhagic stroke, the time-critical detection of CT findings associated with ischemic stroke underlines how timely and qualified interpretation of CT scans is a cornerstone of adequate patient management
We have developed and evaluated a weakly supervised anomaly detection system based on this principle of learning normal anatomy in head CT and flagging anomalies linked to ischemic stroke as deviations from this norm
Summary
Computed tomography (CT) of the head remains the primary modality in stroke imaging. The CT appearances of brain ischemia vary considerably from obvious defects in chronic ischemia to somewhat less obvious findings such as hyperdense vessel sign (HVS) as a surrogate for occluding thrombus. Beyond the mere exclusion of hemorrhagic stroke, the time-critical detection of CT findings associated with ischemic stroke underlines how timely and qualified interpretation of CT scans is a cornerstone of adequate patient management. ML tools can learn through supervised or unsupervised training, with the latter not needing explicit labels for each of the classes (i.e. pathologies) it is supposed to detect. Most ML approaches in medical imaging are based on strongly supervised learning with the associated need for labor-intensive pixel-wise image segmentation and the inherent limitation that a system can only detect what it has previously “seen”. Unsupervised or weakly supervised (only requiring global class labels instead of pixel-level segmentations) systems, on the other hand, offer the potential to learn the underlying data distribution and flag pathology if a derivation from this learned norm is found. Pathology is defined as a deviation from an internalized normal reference and the whole spectrum, not just a predefined library, of imaging anomalies could potentially be detected
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call