Big Data on Healthcare Application

Permanent URI for this collection

https://hdl.handle.net/10125/72197

Browse

Now showing 1 - 9 of 9

Warfarin Dose Estimation on High-dimensional and Incomplete Data
(2021-01-05) Wang, Zeyuan; Poon, Josiah; Yang, Jie; Poon, Simon
Warfarin is a widely used oral anticoagulant worldwide. However, due to the complex relationship between individual factors, it is challenging to estimate the optimal warfarin dose to give full play to its ideal efﬁcacy. Currently, there are plenty of studies using machine learning or deep learning techniques to help with the optimal warfarin dose selection. But few of them can resolve missing values and high-dimensional data naturally, that are two main concerns when analyzing clinical real world data. In this work, we propose to regard each patient’s record as a set of observed individual factors, and represent them in an embedding space, that enables our method can learn from the incomplete date directly and avoid the negative impact from the high-dimensional feature set. Then, a novel neural network is proposed to combine the set of embedded vectors non-linearly, that are capable of capturing their correlations and locating the informative ones for prediction. After comparing with the baseline models on the open source data from International Warfarin Pharmacogenetics Consortium, the experimental results demonstrate that our proposed method outperform others by a signiﬁcant margin. After further analyzing the model performance in different dosing subgroups, we can conclude that the proposed method has the high application value in clinical, especially for the patients in high-dose and medium-dose subgroups.
The Application of Image Recognition and Machine Learning to Capture Readings of Traditional Blood Pressure Devices: A Platform to Promote Population Health Management to Prevent Cardiovascular Diseases
(2021-01-05) Lee, Helen W.Y.; Chu, Christopher T.K.; Yiu, Karen K.L.; Tsoi, Kelvin
Digital solutions for Blood Pressure Monitoring (or Telemonitoring) have sprouted in recent years, innovative solutions are often connected to the Internet of Things (IoT), with mobile health (mHealth) platform. However, clinical validity, technology cost and cross-platform data integration remain as the major barriers for the application of these solutions. In this paper, we present an IoT-based and AI-embedded Blood Pressure Telemonitoring (BPT) system, which facilitates home blood pressure monitoring for individuals. The highlights of this system are the machine learning techniques to enable automatic digits recognition, with F1 score of 98.5%; and the cloud-based portal developed for automated data synchronization and risk stratification. Positive feedbacks on trial implementation are received from three clinics. The overall system architecture, development of machine learning model in digit identification and cloud-based telemonitoring are addressed in this paper, alongside the followed implications.
Predicting Unplanned Hospital Readmissions using Patient Level Data
(2021-01-05) Balan U, Mahesh; Gandhi, Meet; Rammohan, Swaminathan
The rate of unplanned hospital readmissions in the US is likely to face a steady rise after 2020. Hence, this issue has received considerable critical attention with the policy makers. Majority of hospitals in the US pay millions of dollars as penalty for readmitting patients within 30 days due to strict norms imposed by the Hospital Readmission Reduction Program. In this study, we develop two novel models: PURE (Predicting Unplanned Readmissions using Embeddings) and Hybrid DeepR, which uses the historical medical events of patients to predict readmissions within 30 days. Both these models are hybrid sequence models that leverage both sequential events (history of events) and static features (like gender, blood pressure) of the patients to mine patterns in the data. Our results are promising, and they benchmark previous results in predicting hospital readmissions. The contributions of this study add to existing literature on healthcare analytics.
Machine Learning Based Diagnostics of Developmental Coordination Disorder using Electroencephalographic Data
(2021-01-05) Buettner, Ricardo; Buechele, Michael; Grimmeisen, Benedikt; Ulrich, Patrick
We report on promising results concerning the fast and accurate diagnosis of developmental coordination disorder (DCD) which heavily impacts the life of affected children with emotional and behavioral issues. Using a machine learning classifier on spectral data of electroencephalography (EEG) recordings and unfolding the traditional frequency bandwidth in a fine-graded equidistant 99-point spectrum we were able to reach an accuracy of over 99.35 percent having only one misclassification. Our machine learning work contributes to healthcare and information systems research. While current diagnostic methods in use are either complicated, time-consuming, or inaccurate, our automated machine-based approach is accurate and reliable. Our results also provide more insights into the relationship between DCD and brain activity which could stimulate future work in medicine.
High-Performance Detection of Corneal Ulceration Using Image Classification with Convolutional Neural Networks
(2021-01-05) Gross, Jan; Breitenbach, Johannes; Baumgartl, Hermann; Buettner, Ricardo
Corneal Ulcer, also known as keratitis, represents the most frequently appearing symptom among corneal diseases, the second leading cause of ocular morbidity worldwide. Consequences such as irreversible eyesight damage or blindness require an innovative approach that enables a distinction to be made between patterns of different ulcer stages to lower the global burden of visual disability. This paper describes a Convolutional Neural Network-based image classification approach that allows the identification of different types of Corneal Ulcers based on fluorescein staining images. With a balanced accuracy of 92.73 percent, our results set a benchmark in distinguishing between general ulcer patterns. Our proposed method is robust against light reflections and allows automated extraction of meaningful features, manifesting a strong practical and theoretical relevance. By identifying Corneal Ulcers at an early stage, we aid reduction of aggravation by preventively applying and consequently tracking the efficacy of adapted medical treatment, which contributes to IT-based healthcare.
Early Diagnosis of Mild Cognitive Impairment with 2-Dimensional Convolutional Neural Network Classification of Magnetic Resonance Images
(2021-01-05) Heising, Luca; Angelopoulos, Spyros
We motivate and implement an Artificial Intelligence (AI) Computer Aided Diagnosis (CAD) framework, to assist clinicians in the early diagnosis of Mild Cognitive Impairment (MCI) and Alzheimer’s Disease (AD). Our framework is based on a Convolutional Neural Network (CNN) trained and tested on functional Magnetic Resonance Images datasets. We contribute to the literature on AI-CAD frameworks for AD by using a 2D CNN for early diagnosis of MCI. Contrary to current efforts, we do not attempt to provide an AI-CAD framework that will replace clinicians, but one that can work in synergy with them. Our framework is cheaper and faster as it relies on small datasets without the need of high-performance computing infrastructures. Our work contributes to the literature on digital transformation of healthcare, health Information Systems, and NeuroIS, while it opens novel avenues for further research on the topic.
A Multi-view Classification Framework for Falls Prediction: Multiple-domain Assessments in Parkinson’s Disease
(2021-01-05) Huang, Xiuyu; Mark, Latt; Matloob, Khushi; Pelicioni, Paulo; Brodie, Matthew; Lord , Stephen; Loy, Clement; Poon, Simon
Falls are one of the most common causes of injury and disability in people with Parkinson’s disease (PD). This study developed an augmented machine learning framework for screening the risk of falling in people with PD using multiple domain assessments. A sample of 109 people with PD (50 fallers and 59 non-fallers) undertook four domains of assessment: disease-specific rating scales, clinical examination measures, physiological assessments, and gait analysis. A multi-view classifying framework was developed from a sequence of procedures and achieved 77.50% average predicting accuracy. The robustness of the multi-view framework was tested by comparing outcomes of three different view selection methods. The developed framework may have implications for clinical decision making, as some of the PD fall risk variables/features may be amenable to treatment. Our results showed that external reliability can be achieved by a simple voting mechanism from multiple, perhaps diverse, perspective consensus.
A Comparative Evaluation of Machine Learning Deployment Approaches in Real Term Environments using the Example of the Detection of Epileptic Seizures
(2021-01-05) Houta, Salima
The detection of epileptic seizures plays an important role in patient safety and therapy. Much research has been done in recent years to detect epileptic seizures using mobile devices. Although the variety of symptoms of certain types of seizures is challenging, progress has been made in identifying certain types of seizures. Machine learning is used in most work in an experimental environment. However, individual and situational aspects play an important role, especially in the detection of epileptic seizures. The improvement of seizure classification through machine learning in everyday life will play an important role in the further development of the technologies in the next few years. The EPItect project is researching the detection of epileptic seizures using an in-ear sensor. A framework for machine learning for the experimental and real environment was developed in the project. In this paper, we provide a comparative evaluation of different approaches to providing machine learning in the real test environment.
Introduction to the Minitrack on Big Data on Healthcare Application
(2021-01-05) Tsoi, Kelvin; Hung, Patrick; Poon, Simon

Browse

Recent Submissions