Automated Alzheimer disease classification using deep learning models with Soft-NMS and improved ResNet50 integration

Background and objective: The accurate classification of Alzheimer disease (AD) from MRI data holds great significance for facilitating early diagnosis and personalized treatment, ultimately leading to improved patient outcomes. To address this challenge, a comprehensive approach is proposed in this study, which integrates advanced deep learning models. Methods: This study introduces an ensemble deep learning model for AD classification, which incorporates SoftNMS into the Faster R-CNN architecture to enhance candidate information merging and improve detection accuracy. Additionally, an improved ResNet50 network is used for feature extraction, effectively extracting richer image features. To process sequence data, the model incorporates the Bidirectional Gated Recurrent Unit (Bi-GRU) as a key component in the feature extraction network. The final classification results are obtained using the enhanced Faster R-CNN. Results: The results demonstrated outstanding classification performance in distinguishing AD cases, with the AD vs CN task achieving the highest accuracy of 98.91%. The model accuracy and precision remained high for AD vs MCI and MCI vs CN groups, indicating its effectiveness in distinguishing different cognitive impairment stages. Additionally, the effectiveness of the approach was confirmed through object detection and feature extraction, demonstrating superiority over existing methods. The high accuracy of the proposed method indicates its potential for early AD diagnosis and personalized treatment. Conclusion: This study presents a novel approach for AD classification using advanced deep learning models. The achieved high precision and accuracy offer promising opportunities for early diagnosis and personalized intervention. Addressing limitations related to data availability and manual annotation is crucial for future research advancements. Integrating multimodal imaging data and conducting longitudinal studies can further enhance our understanding of AD progression