Introduction Stroke remains a leading cause of worldwide morbidity and mortality. Understanding the relation between its clinical and imaging characteristics is crucial for effective management.

Aim of Study To evaluate how time since stroke onset and clinical variables affect the divergence in stroke size segmentation between non-contrast cranial computed tomography(NCCT) and apparent diffusion coefficient(ADC) imaging, using a self-developed artificial intelligence(AI) software in patients with acute ischemic anterior circulation stroke within 24 hours from symptom onset.

Methods An anonymized cohort of patients diagnosed with ischemic stroke, in whom brain magnetic resonance imaging(MRI), diffusion-weighted imaging(DWI), ADC, and NCCT were performed, with manual segmentation carried out by two neuro-interventional radiologists using the MRICroGL software to outline the ischemic area in NCCT and MRI. Divergence was established as a lesion visible on MRI but not on NCCT.

Results Seventy-seven patients with a median age of 77 years (IQR:64-85) were included, with a predominance of female sex(52%). None had a previous history of ischemic stroke, and 69% of the patients had comorbidities, arterial hypertension(62%) being the most prevalent. Lower stroke size volume in each method showed higher volume between the first 5 hours since stroke onset. Spearman’s Rho correlations showed significant association: infarction size in both modalities(0.402), NIHSS at admission and mRS at discharge(0.432), and NCCT infarction volume and segmentation divergence( -0.884).

Conclusion Stroke size and progression time are crucial for enhancing diagnostic accuracy in AI training. NCCT smaller lesions showed more divergence with MRI. Beyond 5 hours from stroke onset, we found no significant difference between both images.