Search

Alzheimer's disease (AD) is a degenerative neurological ailment that begins with memory loss and ultimately leads to a total loss of mental capacity. Researchers are interested in using magnetic resonance imaging (MRI) and positron emission tomography (PET) to find people with mild cognitive impairment (MCI), which is a stage before Alzheimer's disease (AD). Significant hippocampal loss and temporal lobe atrophy characterize the transition from MCI to AD, which can be visualized using T1-W structural MRI. PET visualizes brain glucose metabolism, which indicates neuronal activity, making it a viable neuroimaging method for AD diagnosis. The extraction and fusion of structural and metabolite information about brain alterations contained in multimodal data is crucial for achieving an appropriate classification result. Therefore, in this work a new end-to-end coupled-GAN (CGAN) architecture is introduced. The proposed CGANC network consists of two sub-models: a CGAN for extraction of fused features from multimodal data, and a CNN classifier to classify these features. The proposed CGAN model is trained to encode MRI and PET images into a shared latent space. The fused features are extracted from this shared latent space and then are classified according to particular stage of AD. In order to test the effectiveness of the suggested approach, experiments are done on the publicly available ADNI dataset and compared with state-of-the-art methods. The proposed method's source code will be made freely available at https://github.com/ChandrajitChoudhury/CGAN-AD. • MRI and PET images are fused to utilize structural and metabolic features for AD diagnosis. • An adversarial learning-based method is proposed to extract fused features. • A coupled adversarial subnetwork and a classification subnetwork have been designed. • Experiments are done on the publicly available ADNI dataset. [ABSTRACT FROM AUTHOR]

Availability of high-quality large annotated data is indeed a significant challenge in healthcare. In addition, privacy concerns and data-sharing restrictions often hinder access to large and diverse medical image datasets. To reduce the requirement for annotated training data, self-supervised pre-training strategies on nonannotated data have been extensively used, whereas collaborative algorithm training without the need to exchange the underlying data. In this paper, we introduce a novel federated learning-based self-supervised spatial–temporal transformer's fusion (SSFL) for cardiovascular image segmentation. The integration of spatial–temporal swin transformer is used to extract the features from 3D SAX multiple phases (full cycle of cardiac heart). An efficient self-supervised contrastive framework consisting of a spatial–temporal transformer network with 25 encoders is used to model the temporal features. The spatial and temporal features are fused and forwarded to the decoder for cardiac heart segmentation using cine MRI images. To further improve segmentation, we use an attention-based unpaired GAN model to map or transfer the style from ACDC to M&Ms and use synthetically generated volumes in the proposed self-supervised approach. Experiments with three different cardiovascular image segmentation tasks, such as segmentation of the right ventricle, left ventricle, and myocardium, showed significant improvement compared to the state-of-the-art segmentation framework. • Efficient self-supervised contrastive framework consisting of a spatial–temporal transformer's fusion. • Extracted features from multiple 3D SAX phases (full cardiac heart cycle) for 4D medical image segmentation. • We used 25 encoders to model the temporal features, and passed them to the single decoder. • Uses an attention-based unpaired GAN model to map or transfer the style from ACDC to M&Ms. • Deployed proposed segmentation model in a federated environment. [ABSTRACT FROM AUTHOR]