Big data promising for optimal triage and management of stroke patients
A team of statisticians, neurosurgeons and radiologists from the University of Hong Kong (HKU) and the Hospital Authority has designed a stroke algorithm created from big data that will allow optimal triage and management of patients suspected to have large vessel occlusion (LVO) stroke.
The territory-wide study, conducted from January to June 2018, enrolled 324 ischaemic stroke patients from whom clinical data, CT scans and CT angiograms (CTA) were collected and analyzed to create an algorithmic model aimed to triage and manage patients suspected of having LVO stroke more rapidly.
“A study we previously conducted has shown that the incidence of LVO stroke is 12.5 per 100,000 persons per year [95 percent confidence interval, 11.7 to 13.4], and that a minute of delay in treatment would mean a loss of 1.9 million neurons. This necessitates effective triage so that patients can receive treatment such as endovascular thrombectomy in a timely manner,” commented investigator Dr Anderson Tsang from the Division of Neurosurgery, HKU. [Int J Stroke 2019, doi: 10.1177/1747493019830585]
“The current practice in Hong Kong is that patients with suspected LVO stroke are brought to the closest hospital, with CT brain and thrombolysis performed, before they are transferred to a facility with CTA and thrombectomy capability. The whole process takes around 2.5 hours to complete,” said Tsang.
“The stroke algorithm we designed would allow bypassing the first hospital so that the patient can be brought directly to the hospital where CT brain, CTA, thrombolysis and thrombectomy can all be performed. This would take only around 1.5 hours to complete,” he added.
“Combining comprehensive assessment of demographic data, signs and symptoms of patients, with evaluation of hyperdense vessels via brain CT significantly improves the accuracy of diagnosing acute LVO stroke [AUC, 0.896; accuracy, 87 percent; sensitivity, 79.2 percent],” said investigator Dr Neeraj Mahboobani from the Department of Radiology and Imaging, Queen Elizabeth Hospital, Hong Kong.
The study group is also looking at integrating patient demographics and clinical data with hyperdense vessel detection on brain CT and the Alberta Stroke Program Early CT Score (ASPECTS). The ASPECTS score is a 10-point quantitative score used to assess early ischaemic changes on brain CT. [http://www.aspectsinstroke.com/]
Apart from prehospital LVO stroke diagnosis and triage, another potential application of the algorithm is for automatic screening of CT brain images for LVO stroke, which will enable faster identification of candidates likely to proceed to CTA.
“CTA is not routinely available for all LVO stroke patients. The algorithm can significantly shorten the time for making a diagnosis and proceeding to treatment,” said Tsang.
“Looking ahead, the final model would aim to create and utilize an LVO probability score accessible to clinicians, who can liaise to come up with a management decision for patients who need to undergo CTA and receive tissue-type plasminogen activator therapy or thrombectomy,” said Mahboobani.