Abstract Body: Introduction: Subdural hematomas (SDHs) are increasing in prevalence and are one of the most common pathologic processes necessitating neurosurgical intervention. Acute SDHs may require urgent surgery and detection of chronic SDHs is of increasing importance due to recent data demonstrating successful management with middle meningeal artery embolization. The prognosis for SDH patients varies significantly based on the hematoma's size, the patient's age, and the promptness of treatment. Rapid SDH is a fully automated software with FDA clearance for triage and notification of hemispheric SDH from non-enhanced head CT (NCCT) images. We determined the accuracy of Rapid SDH for both acute and chronic SDH compared to gold standard expert readers.

Methods: In this retrospective, multicenter study, performance of the Rapid SDH software was assessed based on the consensus of three expert neuroradiologists who independently reviewed each scan to confirm the presence and subtype of SDH blinded to the software output. Scans with SDH with volumes <1 mL and patients under the age of 21 were excluded.

Results: A total of 313 cases, including a consecutive series of patients that was enriched for SDH, were identified (mean age, 63 years ± SD 22; 189 male, 111 female, 13 unknown). Three cases were excluded from the analysis (one for volume <1 ml, two for scan quality). Of the 310 remaining, 157 were positive for SDH (48 acute, 47 chronic, 44 mixed, and 18 isodense) and 153 were negative for SDH per consensus of 3 expert neuroradiologists. Performance of the software, based on the neuroradiologist gold standard, demonstrated a sensitivity of 92.4% (95% CI: 87.1, 95.6) and a specificity of 98.7% (95% CI: 95.4, 99.6) for the detection of SDH. The sensitivity for chronic SDH (91.5% 95% CI 80.1-0.96.6) did not differ from acute/subacute SDH (92.7% CI 86.3-0.96.3). The median processing time was approximately 45 seconds.

Conclusion: The results confirm that automated imaging analysis using Rapid SDH provides fast and accurate detection of both acute and chronic SDH on NCCT. Use of this software has the potential to expedite SDH diagnosis and treatment.