Computing and Technology

Permanent URI for this communityhttp://repository.kln.ac.lk/handle/123456789/15605

Browse

Filters

2022 - 20235

Settings

Author: search.filters.author.Welhenge, Anuradhi

Search Results

Now showing 1 - 5 of 5
  • Thumbnail Image
    Item
    Fog Computing based Heart Disease Prediction System using Deep Learning for Medical IoT
    (Faculty of Computing and Technology, University of Kelaniya Sri Lanka, 2023) Welhenge, Anuradhi; Welhenge, Chiranthi; Subodhani, Shanika
    Internet of Things (IoT) is used in all areas because of the benefits it is offering. All most anything can be connected to the internet and data created by these devices can be analyzed to predict results. IoT is helpful in the medical field because it can connect the patients with the healthcare professionals, and the healthcare professionals can monitor their patients remotely and analyze their data and take necessary actions. Because of the huge amount of data in IoT systems, cloud services are utilized to store the data. But this is not a feasible option in medical IoT, because the predictions should be available as quickly as possible, since patients’ lives are at risk. Therefore, edge-fog- cloud architecture is used. Fog nodes can be used to analyze data closer to the edge devices, resulting in much faster predictions and the cloud can be used for storage. This paper proposes a novel fog based architecture for medical IoT based on deep learning. Deep learning is used on the fog nodes to make accurate predictions. This study used data collected from heart patients to predict the heart disease to evaluate the system and yielded a good accuracy.
  • Thumbnail Image
    Item
    Effect of Finite Impulse Response Filters on Activities of Daily Living Classification Algorithms
    (Faculty of Computing and Technology, University of Kelaniya Sri Lanka., 2023) Welhenge, Anuradhi; Welhenge, Chiranthi
    With the increasing aging population, improving the healthcare system is an important task in every country. The largest number of hospitalizations of the elderly people is due to falls. Therefore, many researchers have come up with different fall detection mechanisms. Improving the accuracy of these algorithms is an important task. This paper focuses on the use of Finite Impulse Response filters to improve the accuracy.
  • Thumbnail Image
    Item
    Fog computing based ultrasound nerve segmentation system using deep learning for mIoT
    (2022) Welhenge, Anuradhi
    Internet of Things is an ever expanding field and applications can be used for medical field. Patient monitoring and diagnosis can be done with the help of IoT and the problems of storing large amount of data can be solved by using cloud computing. However, when transmitting large amount of data through the network, the latency will be impacted. This can be eliminated by introducing a fog layer for the processing of data and processed data later can be stored in the cloud. This study proposes a novel architecture for a hospital ultrasound system and deep learning algorithm is used for the nerve segmentation and a good accuracy is achieved.
  • Thumbnail Image
    Item
    Deep learning based breast cancer detection system using fog computing
    (Journal of Discrete Mathematical Sciences & Cryptography, 2022) Welhenge, Anuradhi
    Among the different types of cancers, more women are suffering from breast cancer. Breast cancer can be identified by mammograms or using ultrasounds. Early detection of the cancer can be used to minimize the complexities the women will face. Deep learning based techniques such as convolutional neural networks (CNN) are used to detect the cancer from mammograms or ultrasound scans. In this study, VGGNet based CNN is used to detect the cancer cells. A novel architecture for collecting, processing and storing of patient data is proposed in this study involving a fog layer. This study achieved a high accuracy, sensitivity and specificity compared to previous studies.
  • Thumbnail Image
    Item
    Blood Pressure Estimation from Photoplethysmography with Motion Artifacts using Long Short Term Memory Network
    (Journal of Biomimetics, Biomaterials and Biomedical Engineering (Volume 54), 2022) Welhenge, Anuradhi; Taparugssanagorn, Attaphongse
    Continuous measurement of the Blood Pressure (BP) is important in hypertensive patientsand elderly population. Traditional cuff based methods are difficult to use since it is uncomfortable towear a cuff throughout the day. A more suitable method is to estimate the BP using the Photoplethysmography(PPG) signal. However, it is difficult to estimate a BP when the PPG is corrupted withMotion Artifacts (MAs). In this paper, Long Short Term Memory (LSTM) an extension of RecurrentNeural Networks (RNN) is used used to improve the accuracy of the estimation of the BP from thecorrupted PPG. It shows that an accuracy of 97.86 is achieved.