https://doi.org/10.1140/epjp/s13360-026-07577-w
Regular Article
Region vision transformer-based dilated densenet for AD classification with the aid of heuristic algorithm
1
Department of Artificial Intelligence and Data Science, Easwari Engineering College, 600089, Ramapuram, Chennai, Tamil Nadu, India
2
Department of Computer Science and Engineering, SRM Institute of Science and Technology, 621105, Tiruchirappalli, Tamil Nadu, India
3
Department of Computer Science and Engineering, National Institute of Technology, 620015, Tiruchirappalli, Tamil Nadu, India
a
This email address is being protected from spambots. You need JavaScript enabled to view it.
Received:
3
November
2025
Accepted:
15
March
2026
Published online:
9
April
2026
Abstract
Alzheimer’s disease (AD) is an increasing neurological disease, which is progressively impairs cognitive function, memory, and daily activities, commonly associated with aging populations. AD is the major reason for dementia, and its occurrence is steadily increasing with the aging global population. Recent advances in diagnostic techniques, including neuroimaging (such as magnetic resonance imaging (MRI) and positron emission tomography (PET)), biomarker analysis, and genetic profiling, have significantly improved our ability to identify early-stage AD. Nonetheless, the challenge remains in accurately classifying the onset of the disease before clinical symptoms become evident. Artificial intelligence (AI) models have shown promising tool for classifying AD risk based on diverse data sources like clinical, genetic, neuroimaging, and cognitive assessment. Despite advancements in AI-driven approaches, accurate and early detection remains challenging, especially in differentiating overlapping disease region. The interpretability and generalizability of the approaches are improved in the integration of AI-driven models in AD detection. Hence, here, an innovative AI-driven deep learning model is presented for the early-stage AD classification, which combines the cutting-edge multimodal data fusion with an optimization strategy. Initially, from the public datasets, the necessary images are gathered for performing AD classification. Further, these images are given into the image preprocessing phase, where the N4 optimal bias correction (N4OBC) with median filtering models is used. The N4OBC is used for removing the artifacts, while the median filtering is utilized for eliminating the unwanted noise from the input images. For getting better preprocessing results, the parameters from N4OBC are optimized by the Updated Finest Search Member-based Intellectual Swarm Bipolar Algorithm (UFSM-ISBA). Subsequently, the preprocessed images were sending to a segmentation phase, where Dense-Inception UNet + + (DIUNet + +) model is employed for segmenting the abnormalities. The segmented images were then given to AD classification phase, where Region Vision Transformer-based Dilated DenseNet (RViT-DDNet) is leveraged to classify the AD. Classified results support an early detection of AD, thereby preventing severe damage and assisting physicians in developing accurate early treatment plans. Finally, the designed AD classification model’s effectiveness is analyzed and compared with the classical models to guarantee the implemented model’s robustness and the results prove that the accuracy of our approach is 94.6% for considering the linear activation function.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to Società Italiana di Fisica and Springer-Verlag GmbH Germany, part of Springer Nature 2026
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
