Alzheimer's disease prediction using 3D-CNNs: Intelligent processing of neuroimaging data.
Authors
Affiliations (8)
Affiliations (8)
- IRC for Finance and Digital Economy, KFUPM Business School, King Fahd University of Petroleum & Minerals, Dhahran 31261, Saudi Arabia.
- Department of Computer Science, University of Science and Technology Bannu, 28100, Pakistan.
- Department of Information Management, National Yunlin University of Science and Technology, 123 University Road, Section 3, Douliou, Yunlin, 64002, Taiwan.
- Hourani Center for Applied Scientific Research, Al-Ahliyya Amman University, Jordan; School of Computing, Skyline University College, University City Sharjah, 1797, Sharjah, UAE.
- Department of Information Systems, College of Computer and Information Sciences, Princess Nourah bint Abdulrahman University, P.O.Box 84428, Riyadh 11671, Saudi Arabia.
- Health Services Academy, Govt of Pakistan, Chak Shahzad, Islamabad, Pakistan.
- Department of Computer Engineering, Gachon University, Seongnam 13120, Republic of Korea. Electronic address: [email protected].
- Department of Computer Science, College of Computer Engineering and Sciences, Prince Sattam Bin Abdulaziz University, Alkharj 11942, Saudi Arabia.
Abstract
Alzheimer's disease (AD) is a severe neurological illness that demolishes memory and brain functioning. This disease affects an individual's capacity to work, think, and behave. The proportion of individuals suffering from AD is rapidly increasing. It flatters a leading cause of disability and impacts millions of people worldwide. Early detection reduces disease expansion, provides more effective therapies, and leads to better results. However, predicting AD at an early stage is complex since its clinical symptoms match with normal aging, mild cognitive impairment (MCI), and neurodegenerative disorders. Prior studies indicate that early diagnosis is improved by the utilization of magnetic resonance imaging (MRI). However, MRI data is scarce, noisy, and extremely diverse among scanners and patient populations. The 2D CNNs analyze 3D data slices separately, resulting in a loss of inter-slice information and contextual coherence required to detect subtle and diffuse brain alterations. This study offered a novel 3Dimensional-Convolutional Neural Network (3D-CNN) and intelligent preprocessing pipeline for AD prediction. This work uses an intelligent frame selection and 3D dilated convolutions mechanism to recognize the most informative slices associated with AD disease. This enabled the model to capture subtle and diffuse structural changes across the brain visible in MRI scans. The proposed model examined brain structures by recognizing small volumetric changes associated with AD and acquiring spatial hierarchies within MRI data. After conducting various experiments, we observed that the proposed 3D-CNNs are highly proficient in capturing early brain changes. To validate the model's performance, a benchmark dataset called AD Neuroimaging Initiative (ADNI) is used and achieves a maximum accuracy of 92.89 %, outperforming state-of-the-art approaches.