AI Based Application Allows Remote Neurological Evaluation of Stroke Patients: An Israeli-European Validation Study. Rambam, Haifa, Israel and Vel d’Hebron, Barcelona, Spain

Background: CVAid Flow (CVAid LTD. Tel-Aviv, Israel) is a novel remote, AI decision-support tool for stroke diagnosis and patient management based on mobile devices.

Aim: To validate “CVAid Flow” system in two comprehensive stroke centers.

Methods: A validation study was conducted during one-year period. 64 patients and11 healthy volunteers were recorded by the system. The software processed the video file presenting comprehensive neurological exam to a remote neurologist. For each patient, 2 evaluations were conducted by 2 different certified stroke neurologists, one at bed side, used as ground truth, and the other through the system, blinded to each other`s assessment.

The primary outcome was inter-rater reliability regarding the neurological severity level among participants measured by intraclass correlation coefficient (ICC). We further calculated ICC adjusted for age, gender and acute vs. subacute.

To detect stroke symptoms CVAid also performed a completely automated AI analysis of facial features.

Results: Overall correlation between the two evaluations was 0.906 (p<0.001). Acute patients showed higher correlation than non acute (0.957, p

There were no significant gender related differences, however patients older than 65 years showed significant stronger correlations than younger patients.

The CVAId facial analysis system showed an accuracy of 87% in stroke symptom detection with sensitivity 95% and Specificity 80%.

Conclusions: “CVAid Flow” efficiently allowed remote neurological evaluation of stroke patients. The automated algorithm was able to accurately triage healthy volunteers from patients suffering a stroke. Further development of additional automated analysis will increase the system accuracy.