Data Science and Intelligent Computing Techniques
Deep Multi-Modal Fusion of Clinical and Non – Clinical Data using Early Submission for Enhanced Kidney Disease Prediction
Authors: Suhasini Tatiparti and Tatiparti B Prasad Reddy
Publishing Date: 09-12-2023
ISBN: 978-81-955020-2-8
Abstract
Due to the shortage of nephrologists and the scarcity of diagnostic labs in rural areas, chronic kidney disease (CKD) has recently emerged as a major health problem. In order to detect illness from clinical and radiological pictures, automated diagnostic models are necessary. Previous research on CKD diagnosis has mostly focused on the independent use of clinical data and CT scans to create AI algorithms. Incorporating clinical data with Computed Tomography (CT) images, this research seeks to create a Deep Multimodal Fusion approach with a late fusion mechanism for diagnosis. Clinical data were used in the tests with CT scans of the kidney to ensure accuracy. Clinical data and image-extracted features are kept separate in the proposed model by a process termed late fusion. The model scored 99 accuracy, 98.6 recall, 97.8 precision, and 98.2 F-score, showing that combining clinical data with the CTD increases diagnosis accuracy to that of a human expert. An additional method of verification was comparing the proposed system's results with those of a human expert. In addition, the findings validate the feasibility of the proposed approach as a diagnostic aid for chronic kidney disease (CKD) for medical professionals.
Keywords
Chronic kidney disease, CT, Precision, Accuracy, Precision, F-score
Cite as
Suhasini Tatiparti and Tatiparti B Prasad Reddy, "Deep Multi-Modal Fusion of Clinical and Non – Clinical Data using Early Submission for Enhanced Kidney Disease Prediction", In: Satyasai Jagannath Nanda and Rajendra Prasad Yadav (eds), Data Science and Intelligent Computing Techniques, SCRS, India, 2023, pp. 711-721. https://doi.org/10.56155/978-81-955020-2-8-63