Page 190 - Hospital Authority Convention 2018

Parallel Sessions
HOSPITAL AUTHORITY CONVENTION 2018

PS12.1 Big Data Analytic 14:30 Room 423 & 424

Rapid Automated Evaluation of Computed Tomography Brain Scan and Prediction for Treatment Needs in Acute
Ischaemic Stroke – A Collaborative “Big Data” Approach
1
2
Tsang ACO , Yu PLH , Leung GKK 1
2
1 Division of Neurosurgery, Department of Surgery, Li Ka Shing Faculty of Medicine, Department of Statistics and Actuarial
Sciences, Faculty of Science, The University of Hong Kong, Hong Kong
Endovascular mechanical thrombectomy is now an established treatment for acute ischemic stroke caused by large vessel
occlusion (LVO). As the majority of ischaemic stroke patients would first undergo plain Computed Tomography (CT) brain
scan, it is imperative that signs of LVO on plain CT scans can be identified quickly and reliably so to initiate confirmatory
investigations, clinical referral and proper treatment. The objective of the present study is to develop a novel, rapid and
automated computer algorithm capable of detecting and predicting signs and likelihood of LVO. The ultimate goal is to
generate an effective, reliable and locally relevant platform for triaging acute stroke patients.
Clinical and imaging data are provided by the Hospital Authority. All patients with stroke-related admissions between 2016
and 2017 are included. CT scans of 300 patients are subject to initial screening by a clinical team of Specialist Neurosurgeons
to determine the “ground truth” (i.e., the presence of absence of LVO in individual patients). Correlations are made with the
clinical course and outcome of the patients in order to determine the actual severity of stroke and the eligibility/need for
advanced treatment at the time of admission.

Relevant predictors of LVO including background risk factors, presenting symptoms, neurological assessments and dense
MCA sign on CT were identified and fed to the machine learning algorithm for predicting the likelihood of LVO; validity testing
will be performed on new dataset to determine sensitivity and specificity, taking into account other clinical parameters and
variable. Here we present the developmental process of this novel and landmark collaborative platform as well as some of our
preliminary findings.

PS12.2 Big Data Analytic 14:30 Room 423 & 424

Unlocking Evidence through Healthcare Text Mining
1
Poon J , Lau A 2
1
School of Information Technologies, University of Sydney, Australia
Tuesday, 8 May 2018 Text is all around. Text fills up medical reports, consultation notes, and discharge summaries. These important clinician
Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Hong Kong
2
narratives are now routinely stored in Electronic Health Records (eHR); harnessing data by text mining could offer new
opportunities for epidemiological research, clinical decision support, meta-analysis and observational research though
advanced data analytics. We know that free-text narrative is invaluable clinical data, but its unstructured nature remains a key
barrier for wide spread use in evidence based medicine.
In this presentation, we will review recent developments in applying text-mining in medical research, including automated
harvesting of clinical concepts and events, clinical coding, and enhancing the accuracy and quality of other structured
clinical data. We shall explore the potential to apply text mining to support clinical studies. We shall also highlight the main
challenges, discuss our clinical-data scientist team approach in data and text mining, and our few attempts in free-text
extraction using data captured for stroke outcomes and colonoscopy safety in the Hospital Authority Clinical Management
System eHR.

188