Cerebriu Apollo

Efficacy

1. Retrospective study of AI tool's ability to identify ischemia in brain MRI. Peer reviewed paper in European Journal of Radiology.AI tool shows 89 % sensitivity (95 % CI: 85 %–91 %) and 90 % specificity (95 % CI: 87 %–92 %) for acute/subacute ischemia.AI’s sensitivity declines with smaller lesions.AI’s specificity declines with artifact presence in scans.No evidence of age, gender, or comorbidity bias in AI tool.
   ‍
2. Retrospective cohort study (OCEBM 3): focusing on the development and validation of an automatic method for assessing DWI/FLAIR mismatch in stroke patients using MRI scans. The goal is to automate the segmentation of parenchymal hyperintensities on FLAIR imaging and create a continuous, rather than binary, DWI/FLAIR mismatch assessment to determine eligibility for recombinant tissue-type plasminogen activator (r-tPA) treatment in ischemic stroke patients, particularly in cases with unknown onset times. Main FindingsPerformance of the Automatic Method: The automatic segmentation method for FLAIR hyperintensities showed strong agreement with neuro-radiologist assessments, achieving a DICE score of 0.820 compared to the radiologists' score of 0.856. This demonstrates that the automated method performs comparably to human experts.
   
   The method was robust to variations in parameters, meaning it should generalize well across different datasets and scanner types.Inter-Rater Agreement: The DICE score, which measures the overlap between different segmentations, was comparable between the automatic method and the neuro-radiologists, indicating that the automated tool is as reliable as human assessments.
   
   Correlation with Clinical Assessments: The automated DWI/FLAIR mismatch ratio and intensity measures correlated well with clinical DWI/FLAIR mismatch assessments made in real-world clinical settings, showing statistical significance (PBCC = 0.60, p < 0.00002).A t-test also showed significant differences between mismatch and no-mismatch groups in both DWI/FLAIR mismatch ratios and intensity measures.
   
   Time Efficiency: The method runs efficiently, taking less than one minute in the worst-case scenario to produce results after image acquisition, making it practical for clinical use in real-time settings.
   ‍
3. Cerebriu Apollo trained on 1300 MRI heads of acute cerebral infarcts. Test vs experienced radiologist. This study involved 30 subjects, sensitivity 94%, specificity 83%. One false negatives and 2 false positives. OCEBM 4 as tiny sample size will be subject to a lot of bias.
   ‍
4. On an independent dataset of 88 scans, the turn around time from scanning a sequence to reporting results back to the hospital system was less than 60 seconds. The specificity and sensitivity for detection was for tumour 95% (88-99%) and 78% (52-94%), and for infarcts 75% (63-85%) and 100% (83-100%). In a study with simulated protocols, on an average, 1.25 fewer sequences were acquired per patient and an overall 0.23 specialised sequences were missed for patients with pathology.