A Novel and Adaptive Evaluation Mechanism for Deep Learning Models in Medical Imaging and Disease Recognition

Loukil, Zainab ORCID: 0000-0003-2731-7051, Ali Mirza, Qublai Khan ORCID: 0000-0003-3403-2935 and Sayers, William ORCID: 0000-0003-1677-4409 (2024) A Novel and Adaptive Evaluation Mechanism for Deep Learning Models in Medical Imaging and Disease Recognition. In: 2023 10th International Conference on Future Internet of Things and Cloud (FiCloud). IEEE, pp. 270-277. ISBN 9798350316353

Text 13776 Loukil et al (2024) A Novel and Adaptive Evaluation Mechanism for Deep Learning Models Matrix_conference_paper_Accepted.pdf - Accepted Version Restricted to Repository staff only until 13 August 2025. (Publisher Embargo). Available under License All Rights Reserved. Download (933kB)

Abstract

Recent advancements in deep learning (DL) and machine learning (ML) algorithms have led to their extensive use in various fields, including healthcare, finance, image processing, etc. However, selecting the most appropriate evaluation metrics for these algorithms remains a challenge. In this paper, we propose a novel evaluation metrics mechanism that takes into account the problem domain and the specific application area. The proposed mechanism involves randomly assigning weights to each metric for each dimension, applying a correlation operation to measure the degree to which these variables are related, and repeating this process for all stages. The resulting matrix is then used to calculate the final Performance Measurement Matrix (PMM) vector that reflects the most optimal measurement metrics. Our proposed mechanism provides a systematic and objective approach to selecting evaluation metrics for DL and ML algorithms, and can be applied to a wide range of applications. We demonstrate the effectiveness of our mechanism using a case study on medical image processing applications.

Item Type:	Book Section
Additional Information:	Presented at the 2023 10th International Conference on Future Internet of Things and Cloud (FiCloud) Marrakesh, Morocco 14 August 2023 - 16 August 2023
Uncontrolled Keywords:	Measurement; Machine learning algorithms; Medical services; Sensitivity and specificity; Reproducibility of results; Safety; Reliability; Evaluation metrics; Perormance evaluation; Model evaluation mechanism; Deep Learning; Measurement matrix
Subjects:	Q Science > QA Mathematics > QA76 Computer software
Divisions:	Schools and Research Institutes > School of Business, Computing and Social Sciences
Research Priority Areas:	Applied Business & Technology
Depositing User:	Susan Turner
Date Deposited:	21 Feb 2024 13:05
Last Modified:	21 Feb 2024 15:30
URI:	https://eprints.glos.ac.uk/id/eprint/13776

University Staff: Request a correction | Repository Editors: Update this record