4 sonuçlar
Arama Sonuçları
Listeleniyor 1 - 4 / 4
Yayın ANN activation function estimators for homomorphic encrypted inference(Institute of Electrical and Electronics Engineers Inc., 2025-06-13) Harb, Mhd Raja Abou; Çeliktaş, BarışHomomorphic Encryption (HE) enables secure computations on encrypted data, facilitating machine learning inference in sensitive environments such as healthcare and finance. However, efficiently handling non-linear activation functions, specifically Sigmoid and Tanh, remains a significant computational challenge for encrypted inference using Artificial Neural Networks (ANNs). This study introduces a lightweight, ANN-based estimator designed to accurately approximate activation functions under homomorphic encryption. Unlike traditional polynomial and piecewise linear approximations, the proposed ANN estimators achieve superior accuracy with lower computational overhead associated with bootstrapping or high-degree polynomial techniques. These estimators are trained on plaintext data and seamlessly integrated into encrypted inference pipelines, significantly outperforming conventional methods. Experimental evaluations demonstrate notable improvements, with ANN estimators enhancing accuracy by approximately 2% for Sigmoid and up to 73% for Tanh functions, improving F1-scores by approximately 2% for Sigmoid and up to 88% for Tanh, and markedly reducing Mean Square Error (MSE) by up to 96% compared to polynomial approximations. The ANN estimator achieves an accuracy of 97.70% and an AUC of 0.9997 when integrated into a CNN architecture on the MNIST dataset, and an accuracy of 85.25% with an AUC of 0.9459 on the UCI Heart Disease dataset during ciphertext inference. These results underscore the estimator’s practical effectiveness and computational feasibility, making it suitable for secure and efficient ANN inference in encrypted environments.Yayın Retinal disease classification from bimodal OCT and OCTA using a CNN-ViT hybrid architecture(Institute of Electrical and Electronics Engineers Inc., 2025-09-21) Aydın, Ömer Faruk; Tek, Faik Boray; Turkan, YaseminRetinal diseases are the leading cause of vision impairment and blindness worldwide. Early and accurate diagnosis is critical for effective treatment, and recent advances in imaging technologies such as Optical Coherence Tomography (OCT) and OCT Angiography (OCTA), have enabled detailed visualization of the retinal structure and vasculature. By leveraging these modalities, this study proposes an advanced deep learning architecture called MultiModalNet for automated multi-class retinal disease classification. MultiModalNet employs a dual-branch design, where OCTA projection maps are processed through a ResNet101 encoder, and cross-sectional slices from the OCT volume (B-scans) are analyzed using a Vision Transformer (ViT-Large). The extracted features from both branches were fused and passed through the fully connected layers for the final classification. Evaluated on the 3-class OCTA-500 dataset, which includes Age-related Macular Degeneration (AMD), Diabetic Retinopathy (DR), and Normal cases, the proposed model achieved state-of-the-art classification accuracy of 94.59 percent, significantly o utperforming single-modality baselines. This result highlights the effectiveness of integrating vascular and structural information to improve the diagnostic performance. The findings suggest that hybrid multi-modal deep learning approaches can play a transformative role in computer-aided ophthalmology, enhancing both clinical decision-making and screening workflows.Yayın Self-supervised learning of 3D structure from 2D OCT slices for retinal disease diagnosis on UK biobank scans(Institute of Electrical and Electronics Engineers Inc., 2025-09-21) Nazlı, Muhammet Serdar; Turkan, Yasemin; Tek, Faik BorayThis study presents a self-supervised learning framework for retinal disease classification using Optical Coherence Tomography (OCT) scans. To balance the contextual richness of 3D volumes with the computational efficiency of 2D architectures, we introduce a quasi-3D input generation strategy. Each input is constructed by stacking three OCT slices, sampled from channel-specific Gaussian distributions centered on the volume midplane, and arranged in a standard three-channel 2D format compatible with existing pre-trained models. These quasi-3D images are used to pre-train a Vision Transformer (ViT-Base) via a Masked Autoencoder (MAE) with a shared masking pattern, encouraging the model to reconstruct masked regions by encoding anatomical continuity across slices. Pre-training is conducted on 10,000 unlabeled OCT volumes from the UK Biobank. The encoder is then fine-tuned on the OCTA-500 dataset for three-class and four-class retinal disease classification tasks, including macular degeneration and diabetic retinopathy. The model achieves 92.57% accuracy on the three-class task, matching the performance of RETFound while using over 150 times less pre-training data and a smaller backbone.Yayın Secure and interpretable dyslexia detection using homomorphic encryption and SHAP-based explanations(Institute of Electrical and Electronics Engineers Inc., 2025-10-25) Harb, Mhd Raja Abou; Çeliktaş, Barış; Eroğlu, GünetProtecting sensitive healthcare data during machine learning inference is critical, particularly in cloud-based environments. This study addresses the privacy and interpretability challenges in dyslexia detection using Quantitative EEG (QEEG) data. We propose a privacy-preserving framework utilizing Homomorphic Encryption (HE) to securely perform inference with an Artificial Neural Network (ANN). Due to the incompatibility of non-linear activation functions with encrypted arithmetic, we employ a dedicated approximation strategy. To ensure model interpretability without compromising privacy, SHapley Additive exPlanations (SHAP) are computed homomorphically and decrypted client-side. Experimental evaluations demonstrate that the encrypted inference achieves an accuracy of 90.03% and an AUC of 0.8218, reflecting only minor performance degradation compared to plaintext inference. SHAP value comparisons (Spearman correlation = 0.59) validate the reliability of the encrypted explanations. These results confirm that integrating privacy-preserving and explainable AI approaches is feasible for secure, ethical, and compliant healthcare deployments.
