Search

Your search keyword '"Hosseini, S. Maryam"' showing total 13 results
Start Over Author "Hosseini, S. Maryam"
13 results on '"Hosseini, S. Maryam"'

1. Distributed Learning and Inference Systems: A Networking Perspective

Author

Moussa, Hesham G., Akhavain, Arashmid, Hosseini, S. Maryam, and McCormick, Bill

Subjects
Computer Science - Machine Learning, Computer Science - Networking and Internet Architecture
Abstract

Machine learning models have achieved, and in some cases surpassed, human-level performance in various tasks, mainly through centralized training of static models and the use of large models stored in centralized clouds for inference. However, this centralized approach has several drawbacks, including privacy concerns, high storage demands, a single point of failure, and significant computing requirements. These challenges have driven interest in developing alternative decentralized and distributed methods for AI training and inference. Distribution introduces additional complexity, as it requires managing multiple moving parts. To address these complexities and fill a gap in the development of distributed AI systems, this work proposes a novel framework, Data and Dynamics-Aware Inference and Training Networks (DA-ITN). The different components of DA-ITN and their functions are explored, and the associated challenges and research areas are highlighted., Comment: This paper has been submitted to IEEE Network magazine and is still under review

Published
2025

2. Cluster Based Secure Multi-Party Computation in Federated Learning for Histopathology Images

Author

Hosseini, S. Maryam, Sikaroudi, Milad, Babaei, Morteza, and Tizhoosh, H. R.

Subjects
Computer Science - Cryptography and Security, Computer Science - Artificial Intelligence, Computer Science - Machine Learning
Abstract

Federated learning (FL) is a decentralized method enabling hospitals to collaboratively learn a model without sharing private patient data for training. In FL, participant hospitals periodically exchange training results rather than training samples with a central server. However, having access to model parameters or gradients can expose private training data samples. To address this challenge, we adopt secure multiparty computation (SMC) to establish a privacy-preserving federated learning framework. In our proposed method, the hospitals are divided into clusters. After local training, each hospital splits its model weights among other hospitals in the same cluster such that no single hospital can retrieve other hospitals' weights on its own. Then, all hospitals sum up the received weights, sending the results to the central server. Finally, the central server aggregates the results, retrieving the average of models' weights and updating the model without having access to individual hospitals' weights. We conduct experiments on a publicly available repository, The Cancer Genome Atlas (TCGA). We compare the performance of the proposed framework with differential privacy and federated averaging as the baseline. The results reveal that compared to differential privacy, our framework can achieve higher accuracy with no privacy leakage risk at a cost of higher communication overhead., Comment: Accepted at MICCAI 2022 Workshop on Distributed, Collaborative and Federated Learning

Published
2022

12. Class-imbalanced Unsupervised and Semi-Supervised Domain Adaptation for Histopathology Images.

Author

Hosseini SM, Shafique A, Babaie M, and Tizhoosh HR

Subjects
Humans, Male, Prostatic Neoplasms, Pathology, Image Interpretation, Computer-Assisted
Abstract

In dealing with the lack of sufficient annotated data and in contrast to supervised learning, unsupervised, self-supervised, and semi-supervised domain adaptation methods are promising approaches, enabling us to transfer knowledge from rich labeled source domains to different (but related) unlabeled target domains, reducing distribution discrepancy between the source and target domains. However, most existing domain adaptation methods do not consider the imbalanced nature of the real-world data, affecting their performance in practice. We propose to overcome this limitation by proposing a novel domain adaptation approach that includes two modifications to the existing models. Firstly, we leverage the focal loss function in response to class-imbalanced labeled data in the source domain. Secondly, we introduce a novel co-training approach to involve pseudo-labeled target data points in the training process. Experiments show that the proposed model can be effective in transferring knowledge from source to target domain. As an example, we use the classification of prostate cancer images into low-cancerous and high-cancerous regions.

Published
2023
Full Text
View/download PDF

13. Anatomical and visual outcomes of three different scleral buckling techniques.

Author

Banaee T, Hosseini SM, Ghooshkhanei H, Moosavi M, and Khayyatzadeh-Kakhki S

Abstract

Purpose: To compare the anatomical and visual outcomes of three different scleral buckling techniques and to explore the effect of cryotherapy and subretinal fluid drainage (SRFD) on outcomes of surgery., Methods: This retrospective study was performed on 111 eyes of 109 patients undergoing scleral buckling for rhegmatogenous retinal detachments (RRDs) by a single surgeon. Pre-, intra- and postoperative data were retrieved from hospital records., Results: Buckles were radial in 27 (24.3%), circumferential (segmental) in 16 (14.4%) and encircling in 68 (61.3%) eyes. Anatomical and visual results were comparable with all three buckling techniques. Application of cryotherapy, the spot number, and SRFD did not affect anatomical and visual results. The only preoperative factor associated with poorer anatomical results was the presence of multiple retinal breaks (P=0.006). The following preoperative factors affected visual outcomes on univariate analysis: extent of retinal detachment (r=0.417, P=0.011) and relative afferent pupillary defect (r=0.423, P=0.02). Preoperative macular status (attached vs detached) also had a significant effect on visual outcomes (P<0.001). Based on multivariate analysis however, only preoperative macular status was significantly correlated with visual results (P=0.022). Silicone sponges placed for non-encircling surgery were removed due to ocular dysmotility in 4 (3.6%) eyes, cosmetic reasons in 3 (2.7%) cases and extrusion in 2 (1.8%) eyes. One encircling tire was also removed due to extrusion., Conclusion: Surgical technique and performing cryotherapy or SRFD do not seem to influence the anatomical and visual outcomes of scleral buckling. Postoperative complications seem to be more prevalent with non-encircling techniques.

Published
2009

Catalog

Books, media, physical & digital resources
See catalog results