To design a predictive model, based on radiographic measurements of admission non-contrasted head computerized tomography (NCHCT), to differentiate perimesencephalic subarachnoid hemorrhage (pmSAH) from aneurysmal causes.

pmSAH is a predominantly benign pattern of subarachnoid hemorrhage. However, in a minority of pmSAH, a posterior circulation aneurysm is causative. Out of concern for missing an aneurysm, many institutions pursue exhaustive imaging, leading to longer ICU length of stay and a considerable resource burden. Previous efforts to predict the presence of aneurysm relied on qualitative visual inspection of hemorrhage patterns. 

We retrospectively reviewed consecutive patients admitted for suspected aneurysmal SAH (aSAH) to an academic center. Patients with a final diagnosis of pmSAH or posterior circulation aSAH were included. Using NCHCT, thickness (continuous variable) and location of blood in basal cisterns and Sylvian fissures (categorical variables) were compared between groups. Using the statistically significant features, we created a scoring system. Receiver operating characteristics (ROC) were used to measure accuracy of this model in predicting aneurysmal etiology. 

Of 420 SAH cases, we identified 49 with pmSAH and 46 with posterior circulation aSAH. Mean age was 53 years (SD 11) and 46 patients (48%) were female. Blood thickness measurements in the crural and ambient cisterns, interhemispheric and Sylvian fissures, and degree of extension into the Sylvian fissure were all statistically significant (p= <0.001, <0.001, <0.001, <0.001, and <0.001, respectively). Using these significant figures, we developed a 10-point scoring model to predict aneurysmal causes with high accuracy (area under the curve [AUC] 0.99; 95% CI 0.98-1.00; odds ratio per point increase: 10; 95% CI 2.18-46.4).

If externally validated, our predictive model may assist clinicians in the risk-stratification of patients presenting with pmSAH. This model could minimize protracted admissions to intensive care units and reduce healthcare resource utilization and costs.