Modelling and Evaluation of Serial and Parallel Density-Based Clustering for Acute Respiratory Distress Syndrome
Year: 2019
Type: dissertation
Abstract: Acute Respiratory Distress Syndrome (ARDS) is a highly heterogeneous condition that affects a relatively low number of mechanically-ventilated intensive care patients, though its mortality rate is quite high, hence the need to develop early diagnosis methods. In this thesis we developed, modelled, and evaluated an approach to extract patient data and evaluate it, perform pre-processing steps, and cluster it using the density-based DBSCAN clustering algorithm. At the same time we made use of available High Performance Computing (HPC) technology in order to parallelise the data manipulation processes. Our implementation showed a significant speed-up in the performance of the highly compute- and data-intensive tasks when using the parallel computing infrastructure as opposed to running them in serial. Additionally, this process was used to assist in selecting the most effective values for the clustering parameters. These results set the path for future implementations of parallel computing in the large scale data extraction and analysis steps necessary for real-time diagnosis in a hospital environment. Furthermore, the pre-processing methods developed in this thesis can easily be adapted to other conditions in the ICU through the use of a HPC-ready Jupyter Notebook, which ultimately makes parallel computing accessible to clinicians without the need for expertise in parallel computing script development.
Source: Skemman
Author: Chadi Barakat
Language: English
Subfield: Health Information Management
Field: Health Professions
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W3080219066
