An Integrated Approach to a Next-Generation Telemedicine and Health Advice System (ANGTHAS) with Artificial Intelligence

  • Shah Sufi Nesar Uddin Siddiqui

Student thesis: Doctoral Thesis


Background: Knowledge engineering is essential for a knowledge-based clinical decision support system for a medical ontology and it assists medical professionals in taking decisions. This thesis presents an investigation into the use of artificial intelligence to support medical decision-making. An AI-enabled health portal is proposed to provide better access to medical advice. A prototype has been built, and the concept has been demonstrated through two investigations: AI for diagnosing Covid-19 and pneumonia, and AI for interpreting X-ray images of bone fractures.
Methodology: A rational approach is used for the original contribution and to understand the research problems with a unique method for each chapter. X-ray and CT images from the Kaggle repository are assessed and downloaded. VGG-16/19, InceptionV3, custom CNN models and various combinations and number of datasets are used for Covid-19, pneumonia and bone fracture ML experiments. Besides, a new CNN model is constructed, and three publicly available x-ray wrist fracture datasets (Mendeley, GRAZPEDWRI- DX, Mura) are combined to increase the model’s performance. Additionally, to build the prototype application, a framework is developed and followed until the finish of the project.
Results: For Covid-19 diagnosis using X-ray images, VGG16 achieved 90% accuracy, 91% sensitivity, and 89% sensitivity. For diagnosis of Covid-19 diagnosis with CT images, VGG16 achieved 100% accuracy, sensitivity, and specificity. For diagnosis of pneumonia with X-ray images, all three models, VGG16, CNN and VGG19, achieved 98% accuracy. VGG19 had the highest sensitivity (100%) and VGG16 had the highest specificity (98%). For wrist fracture diagnosis, the pre-existing and newly built CNN deep learning algorithms did not achieve more than 82% accuracy in identifying fractures when trained on existing public datasets. However, when the newly built CNN was used with a combined dataset, it identified fractures with 98% accuracy, 97% sensitivity, and 97% specificity. The prototype knowledge-based clinical decision support application has been completed, and it is ready for the addition of biomedical information to augment medical opinions.
Discussions and conclusions: The proposed AI-driven technology and digital health care solutions can help people in many areas globally, especially those with disabilities and who live in hard-to-reach areas. Additionally, transferring global real-time public health information with digital medical technology and knowledge-based clinical decision support (KBCDS) increases the capacity to track disease and save lives.
Date of Award28 Feb 2023
Original languageEnglish
Awarding Institution
  • University of Portsmouth
SupervisorAdrian Hopgood (Supervisor), Alice Good (Supervisor) & Alexander Gegov (Supervisor)

Cite this