AI Triaging
Classification of scans with and without suspicions at high precision
Smart Reporting
Productivity gain in reporting scans with "high suspicion"
Build Trust, Measure RoI on AI, Enable Feedback
Cost impact of automating normal scan reporting
Parameter | Description | Hospital A | Hospital B | Hospital C | Hospital D |
---|---|---|---|---|---|
Precision (Normal) | Out of all the cases predicted by Al as Normal, what percentage of them are actually Normal | 99% | 97% | 97% | 96% |
False Discovery Rate(FDR) | Out of all cases segregated by Al as normal, what percentage of them are actually abnormal | 1% | 3% | 3% | 4% |
Reduction in workload | Percentage of all scans that can be reported as normal by Al | 63% | 55% | 37% | 45% |
Error rate (in clinical settings) | Out of all scans predicted by radiologist as normal, what percentage of them were actually abnormal | 8% | 6% | 8% | 11% |
Presented at RSNA 2023 : Segregation of normal chest radiographs from abnormal chest radiographs using DeepTek AI: retrospective and prospective analyses. By Pant, R., Gupte, T., Varma, S., Kulkarni, V., & Kharat, A.
US FDA-cleared CADe software that assists in the analysis of chest X-rays by flagging suspicious regions in the lungs, pleura, heart, and medical hardware.