Article Text

Download PDFPDF
Original research
Yield of diagnostic imaging in atraumatic convexity subarachnoid hemorrhage
  1. Katarina Dakay1,
  2. Ali Mahta1,2,
  3. Shyam Rao1,2,
  4. Michael E Reznik1,2,
  5. Linda C Wendell1,2,
  6. Bradford B Thompson1,2,
  7. N Stevenson Potter1,2,
  8. Ali Saad1,
  9. Chirag D Gandhi3,
  10. Justin Santarelli3,
  11. Fawaz Al-Mufti3,
  12. Brian MacGrory1,
  13. Tina Burton1,
  14. Mahesh V Jayaraman1,2,4,
  15. Ryan A McTaggart1,2,4,
  16. Karen Furie1,
  17. Shadi Yaghi1,5,
  18. Shawna Cutting1
  1. 1 Department of Neurology, Warren Alpert Medical School at Brown University, Providence, Rhode Island, USA
  2. 2 Department of Neurosurgery, Warren Alpert Medical School at Brown University, Providence, Rhode Island, USA
  3. 3 Department of Neurosurgery, Westchester Medical Center at New York Medical College, Valhalla, New York, USA
  4. 4 Department of Diagnostic Imaging, Warren Alpert Medical School at Brown University, Providence, Rhode Island, USA
  5. 5 Department of Neurology, NYU Langone Health, New York, New York, USA
  1. Correspondence to Dr Katarina Dakay, Department of Neurology, Brown University Warren Alpert Medical School, Providence, RI 02903, USA; kdakay{at}


Introduction Atraumatic convexity subarachnoid hemorrhage is a subtype of spontaneous subarachnoid hemorrhage that often presents a diagnostic challenge. Common etiologies include cerebral amyloid angiopathy, vasculopathies, and coagulopathy; however, aneurysm is rare. Given the broad differential of causes of convexity subarachnoid hemorrhage, we assessed the diagnostic yield of common tests and propose a testing strategy.

Methods We performed a single-center retrospective study on consecutive patients with atraumatic convexity subarachnoid hemorrhage over a 2-year period. We obtained and reviewed each patient’s imaging and characterized the frequency with which each test ultimately diagnosed the cause. Additionally, we discuss clinical features of patients with convexity subarachnoid hemorrhage with respect to the mechanism of hemorrhage.

Results We identified 70 patients over the study period (mean (SD) age 64.70 (16.9) years, 35.7% men), of whom 58 patients (82%) had a brain MRI, 57 (81%) had non-invasive vessel imaging, and 27 (38.5%) underwent catheter-based angiography. Diagnoses were made using only non-invasive imaging modalities in 40 patients (57%), while catheter-based angiography confirmed the diagnosis in nine patients (13%). Further clinical history and laboratory testing yielded a diagnosis in an additional 17 patients (24%), while the cause remained unknown in four patients (6%).

Conclusion The etiology of convexity subarachnoid hemorrhage may be diagnosed in most cases via non-invasive imaging and a thorough clinical history. However, catheter angiography should be strongly considered when non-invasive imaging fails to reveal the diagnosis or to better characterize a vascular malformation. Larger prospective studies are needed to validate this algorithm.

  • subarachnoid
  • MRI
  • magnetic resonance angiography
  • hemorrhage
  • angiography

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Atraumatic convexity subarachnoid hemorrhage (cSAH) accounts for approximately 5–7% of all non-traumatic cases of subarachnoid hemorrhage (SAH),1–3 and is defined as isolated hemorrhage in the peripheral sulci without extension into the basal cisterns or Sylvian fissure.4 There are numerous potential causes, including cerebral amyloid angiopathy (CAA), reversible cerebral vasoconstriction syndrome (RCVS), venous sinus thrombosis, and coagulopathy, among others. However, in up to 14–35% of cases no etiology is ever identified.5

Due to its relatively low incidence and heterogenous clinical presentation, there is no clear consensus for diagnostic workup of cSAH. One group proposed a diagnostic approach on the basis of a small retrospective case series, emphasizing the use of brain MRI and magnetic resonance angiography (MRA) and proceeding to catheter-based angiography in cases where diagnostic uncertainty remained.6 We aimed to consider the diagnostic utility of other non-invasive tests such as computed tomographic angiography (CTA), magnetic resonance venography (MRV), and transcranial Doppler ultrasound (TCD).


Study cohort

We retrospectively identified consecutive adult patients with non-traumatic, high convexity, non-aneurysmal SAH admitted to our academic comprehensive stroke center from January 2015 to December 2017. All cases of SAH were identified using the Get With the Guidelines (GWTG) database. We excluded patients with a clear antecedent history of trauma preceding the SAH and those with hemorrhage in the basal or Sylvian cisterns. cSAH was defined as either hyperdensity on CT or hyperintense signal on MRI FLAIR isolated to the sulci without involving the parenchyma, basal cisterns, or Sylvian fissure. Patients with a primary intraparenchymal hemorrhage with subarachnoid extension, iatrogenic SAH (after treatment with thrombolysis or mechanical thrombectomy), and pediatric patients aged <18 were excluded. In patients with cancer, to avoid including leptomeningeal disease which can mimic cSAH on the FLAIR sequence, patients either had to have hyperdensity on initial CT or a hypointensity on the gradient echo sequence in addition to meeting the above inclusion criteria. Patients whose etiology of cSAH was known prior to presentation, such as a known arteriovenous malformation (AVM) or an ischemic stroke readily apparent on CT, were also excluded. Two patients ultimately diagnosed with ischemic stroke on MRI initially presented with non-focal symptoms and a CT unrevealing of their stroke, and were included as they required further diagnostic workup for their cSAH.

Study data

Patient demographics, comorbidities, clinical testing, and other data related to standard clinical stroke care were collected as part of an institutional quality improvement project. We also collected detailed data from patients’ brain MRIs, MRA, CTA, MRV, and catheter-based angiogram. MRI and MRA were performed on a 1.5 T magnet. Data were abstracted from the electronic medical record into a secure password-protected database (REDCap). This study was approved by the Institutional Review Board.

Study adjudication

All cases of cSAH and their likeliest etiology were adjudicated by the attending neurologist as well as the vascular neurology fellow performing data collection. The modified Boston criteria were used to classify patients as having possible or probable CAA,7 and RCVS was diagnosed on the basis of a suggestive clinical history in conjunction with appropriate imaging findings. Patients were classified as having an anticoagulation-related hemorrhage if they were actively taking an anticoagulant medication (with confirmatory laboratory testing as applicable) and had no evidence of an underlying structural brain lesion. Patients were classified as having a coagulopathy-related hemorrhage if they had an underlying diagnosis of coagulopathy on the basis of pertinent laboratory testing.

Statistical analysis

We analyzed the yield of each specific diagnostic test by determining the frequency with which each test resulted in the accurate identification of the underlying cSAH etiology. Based on comparisons of the various yields for each diagnostic test, taking into account its invasiveness and availability, we then developed a suggested algorithm for the diagnostic workup of atraumatic cSAH.


Patient characteristics

We identified 572 patients with SAH during the study period. Of these, 211 were cisternal pattern SAH and were excluded, leaving 361 patients with cSAH during the study period. However, we excluded 121 patients with an antecedent history of head trauma, and an additional 170 patients had a primary intraparenchymal hemorrhage with subarachnoid extension, SAH secondary to a previously known cause such as an AVM or recent ischemic stroke, or an iatrogenic SAH after endovascular or neurosurgical procedure. Our final study cohort therefore included 70 patients who fulfilled our selection criteria.

The mean (SD) age of our cohort was 64.7 (16.9) years and 64.3% were women (table 1). Hypertension was present in 64.3% of patients, diabetes mellitus in 22.9%, prior stroke in 22.9%, and migraine in 13.2%, while 15.7% of patients were on therapeutic anticoagulation at the time of presentation for cSAH.

Table 1

Patient characteristics


The most common presenting symptom was headache, which occurred in 37.1% (26/70) of patients, while encephalopathy was the second most common symptom and occurred in 27.1% (19/70). Focal neurologic deficits were the presenting symptom in 25.7% (18/70) of patients, and seizure was the initial symptom in 8.6% (6/70) of patients. One patient with unrelated symptoms underwent imaging which revealed cSAH as an asymptomatic finding.

Non-invasive vessel imaging

Non-invasive vessel imaging with either CTA or MRA and MRV was performed in a total of 57 (81.4%) patients. Of the 43 patients who had a CTA, five patients had a causative lesion such as venous sinus thrombosis or diffuse vasospasm suggestive of RCVS. Of the 17 patients who had an MRA and MRV, four patients had positive findings of venous sinus thrombosis and one patient had RCVS.


Brain MRI was performed in 82.8% (58/70) of patients, and ultimately led to a diagnosis in 55.2% (32/58) of patients (figure 1). Of note, MRI identified two cases of SAH which were not visible on CT.

MRI diagnoses included probable CAA in 17 patients (10 with superficial siderosis and 7 with cortical microbleeds), posterior-predominant T2 hyperintensities suggestive of posterior reversible encephalopathy syndrome in one patient, small cortical-based ischemic strokes suggestive of RCVS in three patients, cortical vein thrombosis in one patient, underlying malignancy in four patients, underlying ischemic stroke (with hemorrhagic transformation) in two patients, hyperperfusion syndrome after carotid endarterectomy in two patients, hemorrhagic limbic encephalitis in one patient, and cavernoma in one patient.

Figure 1

MRI findings in convexity subarachnoid hemorrhage. CAA, cerebral amyloid angiopathy; PRES, posterior reversible encephalopathy syndrome; RCVS, reversible cerebral vasoconstriction syndrome . 

Catheter-based angiography

Catheter-based angiography was performed in 27 (38.6%) patients, of whom 12 were found to have causative abnormalities and two others were found to have incidental aneurysms. Culprit etiologies included multifocal vasospasm suggestive of RCVS in seven patients (each case was previously missed with non-invasive vessel imaging; figure 2); a pseudoaneurysm adjacent to the cSAH in one patient with endocarditis; a dural arteriovenous fistula not visualized on either CTA or MRI in one patient; and severe intracranial atherosclerosis with pseudo-Moyamoya collateralization in one patient thought to be implicated in their cSAH.

Figure 2

Magnetic resonance angiography of brain in patient with convexity subarachnoid hemorrhage (normal images on left). Subsequent angiogram (right) demonstrated vasoconstriction of the left posterior cerebral artery (blue arrow) and superior cerebellar artery (yellow arrow) suggestive of RCVS.

Transcranial Doppler ultrasound

TCDs were performed in 22 patients and were normal in 20, including eight patients who were ultimately diagnosed with RCVS. Two patients had findings of elevated velocities potentially suggestive of arterial narrowing; however, these proved to be false positives as the cSAH etiologies for these patients were ultimately thought to be an underlying cavernoma and coagulopathy, respectively.


Ultimately, cSAH etiology was identified in 66/70 patients; for four patients the etiology remained unknown despite a thorough workup (table 2). CAA was the most common cause, accounting for 21 cases (30.0%), followed by RCVS in 14 patients (20.0%). Among patients aged >65, CAA unsurprisingly took on a greater share as the likeliest etiology (44.7% (17/38) of patients), while RCVS was the most common cause in patients aged <65 (36.4% (12/33) of patients). We found non-iatrogenic coagulopathy was the third most frequent cause of cSAH in our cohort, accounting for eight cases (three due to leukemia and one each due to heparin-induced thrombocytopenia, cirrhosis, aplastic anemia, idiopathic thrombocytopenia purpura, and cryoglobulinemia). MRI was the diagnostic test most likely to identify the cause; table 3 summarizes the diagnoses by test that confirmed the cause.

Table 2

Etiology of subarachnoid hemorrhage

Table 3

Diagnostic test and etiology of convexity subarachnoid hemorrhage  (cSAH)

Clinical features were also associated with specific cSAH etiologies in our cohort. Notably, 10% of patients (7/70) had a previous history of cSAH; of these, four had CAA, one had endocarditis, and one patient was found to have a dural arteriovenous fistula. Presenting symptoms also appeared to correspond with etiology: patients with CAA were likely to present with a focal neurologic deficit (11/19 patients), while patients with RCVS were likely to present primarily with thunderclap headache (12/16 patients).


The majority of patients (n=45) were discharged home after their hospital stay. Ten patients had died at 90 days post-discharge; of these, only two were aged <60 years. Sepsis was the most common cause of death (n=3); other causes included pulmonary embolism (n=2), subsequent burn or trauma unrelated to cSAH (n=2), intracerebral hemorrhage (n=2), and aspiration pneumonia (n=1).


In our retrospective cohort we confirmed that RCVS and CAA are the most common etiologies of SAH, similar to previous series.8–10 The combination of non-invasive vessel imaging, MRI and, in selected cases, angiography, led to diagnosis of the cause of the cSAH in 94.2% of patients (66/70). These numbers are similar to the 87% diagnostic yield in a previous case series.6

Similar to previous reports, younger patients were more likely to present with thunderclap headache and to be diagnosed with RCVS, and older patients were more likely to present with transient focal neurologic events from CAA.5 11 Of note, the incidence of coagulopathy-associated hemorrhages was higher in our cohort at 17% compared with 3% previously reported.6

The combination of MRI and non-invasive vessel imaging incorporating both an arterial and a venous phase diagnosed about 57% of cases of cSAH in our cohort; in these patients it is possible they may be able to forego catheter angiography depending on the etiology. The most common diagnosis revealed by the MRI was CAA, where cortical microbleeds and superficial siderosis were seen on fast low-angle shot MRI (FLASH). The venous phase of multiphase CTA and MRV were particularly useful in discerning venous sinus thrombosis, but missed several cases of RCVS-associated vasoconstriction likely due to the limitations of evaluating the distal vasculature. Angiography detected seven of these cases, but also discovered two cases of intervenable lesions including one pseudoaneurysm and one dural arteriovenous fistula which was not visible on either non-invasive vessel imaging or MRI brain. The likelihood of angiography revealing the cause of the cSAH was 37%, similar to other series where the diagnostic yield ranged from 23% to 50%1 2 6; however, angiography was typically reserved for patients in whom non-invasive tests were inconclusive. On the basis of these findings, we propose an algorithm for evaluating cSAH patients in figure 3.

Figure 3

A suggested diagnostic approach to convexity subarachnoid hemorrhage (cSAH) based on retrospective data.

Often, however, imaging did not reveal an underlying structural lesion, but rather ruled them out. In these cases, most patients were diagnosed on the basis of their clinical history and laboratory findings, but in a small proportion of patients the cause remained unknown. Many such cases were due to either therapeutic anticoagulation or a non-iatrogenic coagulopathy, often malignancy-related. Notably, a small number of patients in our cohort were diagnosed with RCVS on the basis of thunderclap headache and the presence of a trigger known to induce RCVS, without the characteristic vasoconstriction seen on vessel imaging. This diagnostic dilemma occurs not infrequently in RCVS, where the vessel imaging can lag behind the clinical symptoms.12 13 TCDs, which have been reported in some cases to aid in the diagnosis of vasospasm,11 were not adequate for screening for RCVS in our patient population; however, they were often only performed once, limiting their utility. It is useful to note that the yield of non-invasive vessel imaging in RCVS may be dependent on the timing of the imaging in relation to the onset of symptoms; it has previously been reported that vasoconstriction begins distally and progresses proximally in RCVS.14 Non-invasive vessel imaging can be falsely negative in distal vasoconstriction; hence, it is possible that repeating non-invasive vessel imaging a few days after presentation in cases of RCVS may reveal vasoconstriction when it was initially undetectable. One limitation of our study was that non-invasive vessel imaging was often not repeated in the acute setting. In cases where RCVS is highly suspected, repeating non-invasive vessel imaging may be an alternative to proceeding directly with catheter-based angiography.

Ultimately, the purpose of our study was to study the yield of diagnostic testing in cSAH. Angiography is a resource-intensive and invasive test, and identifying patients who may not require an angiogram is an important but challenging goal. A risk score for identifying patients most likely to have a positive angiography or, conversely, a scoring system for identifying patients in whom angiography is unlikely to be revealing is a possible area of further study. Our case series, while large in relation to many others, is still small and limited by its retrospective nature. Prospective studies with larger sample sizes may make a more quantitative analysis more feasible.

However, our study was limited in that it is a single-center retrospective study; patients were identified through the Get With the Guidelines database. Given the retrospective nature, it is possible that patients with cSAH not coded as subarachnoid hemorrhage were incorrectly excluded from the study. Additionally, not all patients who presented with RCVS-like symptoms received follow-up vessel imaging to evaluate for delayed vasoconstriction which may have increased the likelihood of identifying radiographic vasospasm. One diagnostic test not evaluated in our cohort is high-resolution vessel wall MRI (HRMRI), a non-invasive imaging test designed to evaluate specific vessels of interest in detail. At our institution, MRI and MRA were performed on a 1.5 T magnet and this could limit the evaluation of the distal vasculature. HRMRI has been previously applied to cases of vasoconstriction where there is a diagnostic challenge in differentiating between RCVS and primary CNS vasculitis, and can show vessel wall enhancement in vasculitis and diffuse uniform thickening of the vessel in RCVS.15 16 It is currently uncertain whether HRMRI may be helpful in patients with thunderclap headache and initially negative vessel imaging and angiography, but may be an interesting topic of further study given the difficulty of demonstrating vasoconstriction in these patients.


Our study adds to the findings of previous series of cSAH and reinforces the importance of MRI and non-invasive vessel imaging in the initial workup of cSAH and of angiography in selected patients. Further studies are warranted to evaluate the utility of other imaging modalities in this condition.



  • Contributors KD: study concept, literature review, data collection, manuscript preparation and revision. SY, SC: study concept, critical review, manuscript revision. All other authors provided substantial critical review and revision of the manuscript.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Ethics approval The study was approved by the local Institutional Review Board at the participating center.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement All data are presented in this paper.

  • Presented at AANS/CNS Joint Section Meeting (E-poster), February 4-5 2019, Honolulu, HI.

  • Patient consent for publication Not required.