Global Journal of Computer Science and Technology, C: Software & Data Engineering, Volume 22 Issue 2

Capillary X: A Software Design Pattern for Analyzing Medical Images in Real-time using Deep Learning Maged Abdalla Helmy Abdou α , Paulo Ferreira σ , Eric Jul ρ & Tuyen Trung Truong Ѡ Abstract- Recent advances in digital imaging, e.g., increased number of pixels captured, have meant that the volume of data to be processed and analyzed from these images has also increased. Deep learning algorithms are state-of-the-art for analyzing such images, given their high accuracy when trained with a large data volume of data. Nevertheless, such analysis requires considerable computational power, making such algorithms time- and resource-demanding. Such high demands can be met by using third-party cloud service providers. However, analyzing medical images using such services raises several legal and privacy challenges and do not necessarily provide real-time results. This paper provides a computing architecture that locally and in parallel can analyze medical images in real-time using deep learning thus avoiding the legal and privacy challenges stemming from uploading data to a third-party cloud provider. To make local image processing efficient on modern multi-core processors, we utilize parallel execution to offset the resourceintensive demands of deep neural networks. We focus on a specific medical-industrial case study, namely the quantifying of blood vessels in microcirculation images for which we have developed a working system. It is currently used in an industrial, clinical research setting as part of an e-health application. Our results show that our system is approximately 78% faster than its serial system counterpart and 12% faster than a master-slave parallel system architecture . I. I ntroduction arly attempts to address the problem of running demanding computational algorithms in tightly constrained environments emerged in the 1980s [1]. Performance limitations became apparent with the rise of processing big data using deep learning (DL) techniques because DL requires large amounts of computational power [2], [3]. Such limitations included the under-utilization of the available computing resources to execute processes introducing undesirable delays [4]. These limitations are still prominent when real-time results are desired in tightly constrained environments (i.e., clinical environments). Furthermore, using third-party cloud services to rent computing resources is risky due to General Data Protection Regulation (GDPR) [5]. These regulations effectively limit clinicians to local computing resources, such as laptops and PCs approved for use at hospitals. Author α σ ρ: Department of Informatics University of Oslo, Norway. e-mail : magedaa@uio.no Author Ѡ : Department of Mathematics University of Oslo, Norway. This paper aims to design, implement, and evaluate a software package that can analyze medical images using deep learning in a local environment as to mitigate the risk of breaching GDPR rules while still getting results in real-time. We focus on a specific industrial, medical case study: the quantification of blood vessels in microcirculation images captured by using in-clinic, hand-held cameras with microscope lenses. The quantified value is called capillary density or blood vessel density. This value is of high clinical relevance because the fluctuation of this value can be used as an early marker to indicate an organ failure, and the severity of the change might predict the chances of the patient surviving [6]–[12]. The requirements of our system were established by interviewing a set of medical doctors and surgeons who spent several years in the microcirculation analysis field (associated with ODI Medical AS, a MedTech company responsible for the e- health industrial application). The main requirements for a production-grade system for the quantification of blood vessels analysis captured by a real-time camera are: 1. The system must be able to analyze a microcirculation image (1920x1080) in real-time (one second or less); 2. The system must have low power consumption so that it can be used in battery-powered devices in hospitals; and 3. The system must be built on top of a popular, widely used programming language and framework (e.g., Python and Tensorflow) running on standard hardware. To the best of our knowledge, no previous work on microcirculation analysis reported using parallel frameworks to calculate the capillary density in under 1 second for a frame with a resolution of 1920x1080 on a CPU using deep learning with an accuracy of 85%. The medical doctors proposed this accuracy value to outperform the accuracy achievable by a trained clinician. Previous systems in the literature that achieve a comparable or higher accuracy needed a GPU that is not available in typical low-power computers approved for use in hospitals. The developed system runs in an industrial, clinical environment on a standard low cost E Global Journal of Computer Science and Technology Volume XXII Issue II Version I 13 Year 2022 ( ) C © 2022 Global Journals ∼