Artificial intelligence for detection of intracranial haemorrhage on head computed tomography scans: diagnostic accuracy in Hong Kong.

INTRODUCTION: The use of artificial intelligence (AI) to identify acute intracranial haemorrhage (ICH) on computed tomography (CT) scans may facilitate initial imaging interpretation in the accident and emergency department. However, AI model construction requires a large amount of annotated data for training, and validation with real-world data has been limited. We developed an algorithm using an open-access dataset of CT slices, then assessed its utility in clinical practice by validating its performance on CT scans from our institution. METHODS: Using a publicly available international dataset of >750 000 expert-labelled CT slices, we developed an AI model which determines ICH probability for each CT scan and nominates five potential ICH-positive CT slices for review. We validated the model using retrospective data from 1372 non-contrast head CT scans (84 [6.1%] with ICH) collected at our institution. RESULTS: The model achieved an area under the curve of 0.842 (95% confidence interval=0.791-0.894; P<0.001) for scan-based detection of ICH. A pre-specified probability threshold of ≥50% for the presence of ICH yielded 78.6% accuracy, 73% sensitivity, 79% specificity, 18.6% positive predictive value, and 97.8% negative predictive value. There were 62 true-positive scans and 22 false-negative scans, which could be reduced to six false-negative scans by manual review of model-nominated CT slices. CONCLUSION: Our model exhibited good accuracy in the CT scan-based detection of ICH, considering the low prevalence of ICH in Hong Kong. Model refinement to allow direct localisation of ICH will facilitate the use of AI solutions in clinical practice.

Translesional pressure gradient and leptomeningeal collateral status in symptomatic middle cerebral artery stenosis.

BACKGROUND AND PURPOSE: Leptomeningeal collateral (LMC) status governs the prognosis of large artery occlusive stroke, although factors determining LMC status are not fully elucidated. The aim was to investigate metrics affecting LMC status in such patients by using computational fluid dynamics (CFD) models based on computed tomography angiography (CTA). METHODS: In this cross-sectional study, patients with recent ischaemic stroke or transient ischaemic attack attributed to atherosclerotic M1 middle cerebral artery (MCA) stenosis (50%-99%) were recruited. Demographic, clinical and imaging data of these patients were collected. Ipsilesional LMC status was graded as good or poor by assessing the laterality of anterior and posterior cerebral arteries in CTA. A CFD model based on CTA was constructed to reflect focal hemodynamics in the distal internal carotid artery, M1 MCA and A1 anterior cerebral artery. Pressure gradients were calculated across culprit MCA stenotic lesions in CFD models. Predictors for good LMC status were sought in univariate and multivariate analyses. RESULTS: Amongst the 85 patients enrolled (mean age 61.5 ± 10.9 years), 38 (44.7%) had good ipsilesional LMC status. The mean pressure gradient across MCA lesions was 14.8 ± 18.1 mmHg. Advanced age (P = 0.030) and a larger translesional pressure gradient (P = 0.029) independently predicted good LMCs. A lower fasting blood glucose level also showed a trend for good LMCs (P = 0.058). CONCLUSIONS: Our study suggested a correlation between translesional pressure gradient and maturation of LMCs in intracranial atherosclerotic disease. Further studies with more exquisite and dynamic monitoring of cerebral hemodynamics and LMC evolution are needed to verify the current findings.

Pontine microbleeds and depression in stroke.

OBJECTIVES: Depression is the most common affective disorder following stroke yet the neuroanatomical model of poststroke depression (PSD) remains unclear. This study examined the association between PSD and cerebral microbleeds (CMBs) and hypothesized that CMBs in specific regions would be associated with PSD. METHODS: Of the 4766 patients with first ever or recurrent acute ischemic stroke admitted to the Acute Stroke Unit of the Prince of Wales Hospital between June 2004 and October 2010, 229 met the entry criteria and formed the study sample. Patients with a Geriatric Depression Scale score of 7 or above were classified as having PSD. The presence and location of CMBs were evaluated with magnetic resonance imaging. RESULTS: Compared to the non-PSD group, patients with PSD were more likely to have pontine CMBs (32.0% vs 18.2%; P = .019). The presence of pontine CMBs remained an independent predictor of PSD in the multivariate analysis, with an odds ratio of 2.2 (P = .016). CONCLUSION: The results suggest that pontine CMBs are associated with a higher risk of developing PSD.

An expedited stroke triage pathway: the key to shortening the door-to-needle time in delivery of thrombolysis.

OBJECTIVES: To assess time management of stroke thrombolysis triage and functional outcomes in patients receiving recombinant tissue plasminogen activator for hyperacute stroke, and identify bottlenecks in delivery of the treatment. DESIGN: Prospective study. SETTING: A university teaching hospital in Hong Kong. PATIENTS: Patients with suspected hyperacute stroke referred to the stroke thrombolysis team during October 2008 to September 2009. MAIN OUTCOME MEASURES: Time performance records including door-to-stroke team, door-to-needle, and onset-to-thrombolysis times. Functional outcomes by modified Rankin Scale score at 3 months, and thrombolysis-related complications including haemorrhagic transformations and mortality. RESULTS: During the 12-month period, 95 thrombolysis calls were received; recombinant tissue plasminogen activator was given intravenously to 17 (18%) of the patients and intra-arterially to 11 (12%). The mean (standard deviation) door-to-stroke team and the door-to-needle times for intravenous recombinant tissue plasminogen activator patients were 33 (25) and 80 (25) minutes, respectively; both were about 20 minutes longer than that recommended by the National Institute of Neurological Disorders and Stroke. The mean National Institute of Health Stroke Scale score for patients received intravenous recombinant tissue plasminogen activator was 16 (standard deviation, 7). The mean (standard deviation) onset-to-treatment time was 144 (42) minutes. Nine (53%) patients who received intravenous recombinant tissue plasminogen activator achieved favourable outcomes at 3 months, with a modified Rankin Scale score of 0 to 1. Symptomatic haemorrhage and mortality occurred in one (6%) patient. CONCLUSION: A dedicated stroke triage pathway is essential to ensure efficient and safe delivery of thrombolysis therapy. Improvements in door-to-stroke team time through integration with emergency medicine staff and neuroradiologists may improve thrombolysis eligibility.

MR imaging of nonmalignant polyps and masses of the nasopharynx and sphenoid sinus after radiotherapy for nasopharyngeal carcinoma.

BACKGROUND AND PURPOSE: The development of a new polyp or mass in the radiation field of a previously treated carcinoma is usually an ominous sign of a recurrent cancer, but rarely may it be caused instead by a nonmalignant process. The purpose of this study was to document the MR appearance of unusual nonmalignant polyps or masses (NMPMs) in the nasopharynx and sphenoid sinus arising after radiation treatment of nasopharyngeal carcinoma. MATERIALS AND METHODS: The MR imaging reports of patients undergoing imaging after radiation therapy for nasopharyngeal carcinoma were reviewed retrospectively to identify patients with unusual polyps and masses in the nasopharynx. The MR images of those patients with no evidence of malignancy on biopsy or follow-up were reviewed. RESULTS: The MR imaging reports of 1282 patients were reviewed, and 11 patients (1%) with NMPMs in the nasopharynx or sphenoid sinus were identified. Two patterns were identified: contrast enhancing nasopharyngeal polyps ranging in size from 1 to 5 cm (n = 5) and sphenoid sinus masses consisting of a nonenhancing mass filling a nonexpanded sinus (n = 4) and a heterogeneous enhancing mass expanding the sinus (n = 2). Osteoradionecrosis produced a large defect in the roof of the nasopharynx causing direct communication with the sphenoid sinus (n = 6). Histology revealed granulation tissue in all of the patients with variable amounts of fibrin and inflammatory cells. A direct infective etiology was not proved in any patient. CONCLUSION: NMPMs in the nasopharynx and sphenoid sinus are rare complications after radiation therapy to the skull base, but the radiologist needs to be aware of their appearance so that they can be considered in the differential diagnosis of suspected tumor recurrence.