6 April 2022 in IEEE Access
Diabetic Retinopathy (DR) - a complication developed as a result of heightened blood glucose levels- is deemed as one of the most sight-threatening diseases. Unfortunately, DR screening is manually acquired by an ophthalmologist, a process that can be considered erroneous and timeconsuming. Accordingly, in recent years, automated DR diagnostics have become a focus of research due to the tremendous increase in diabetic patients. Moreover, the recent accomplishments demonstrated by Convolutional Neural Networks (CNN) settle them as state-of-the-art for DR stage identification. This paper proposes a new automatic deep-learning-based approach for severity detection by utilizing a single Color Fundus photograph (CFP). The proposed technique employs DenseNet169’s encoder to construct a visual embedding. Furthermore, Convolutional Block Attention Module (CBAM) is introduced on top of the encoder to reinforce its discriminative power. The model is trained using cross-entropy loss on the Kaggle Asia Pacific Tele-Ophthalmology Society’s (APTOS) dataset. On the binary classification task, we accomplished (97% accuracy - 97% sensitivity - 98.3% specificity - 0.9455, Quadratic Weighted Kappa score (QWK)) compared to the state-of-the-art. Moreover, Our network showed high competency (82% accuracy - 0.888 (QWK)) for severity grading. The major contribution of the proposed framework is that it efficiently grades the severity level of diabetic retinopathy, while reducing the time and space complexity required, which demonstrates it as a promising candidate for autonomous diagnosis.
18 May 2023 in Memories - Materials, Devices, Circuits and Systems
9 March 2026 in IEEE Access
One-bit Compressed Sensing (CS) is a novel approach in signal processing that focuses on the recovery of sparse signals from highly quantized measurements, specifically retaining only the sign information of the measurements. This extreme quantization reduces the data storage and transmission requirements, making it particularly suitable for resource constrained environments. Despite the loss of amplitude information, one-bit compressed sensing leverages advanced algorithms and theoretical frameworks to achieve accurate signal reconstruction. This technique has shown potential in various applications, including low-power sensor networks and medical imaging, where efficient data acquisition is crucial. This work provides a comparative study over various one-bit CS recovery techniques proposed in the literature with respect to measurement matrix structure, noise resilience, convexity, and side information requirement in terms of either noise or sparsity information.We conducted a comprehensive comparison of multiple algorithms on a certain setup, evaluating their performance in terms of Signal-to-Noise Ratio (SNR), Hamming Distance (HD), and Hamming Error (HE) to assess their effectiveness in signal recovery. We extend our validation process for the recovery techniques through an adaptive setup focused on image compression and reconstruction using one-bit compressed sensing, emphasizing their suitability for hardware implementation in resource-constrained IoT hardware applications. Results provide guidance over appropriate algorithm selection for different IoT use cases. Evaluation process is performed with respect to reconstruction accuracy, reconstruction latency, and storage resources. In noiseless setup, the one-bit LP algorithm supersedes in terms of reconstruction accuracy utilizing small storage resources with latency overhead, meanwhile the Epin algorithm represents a better option in terms of reconstruction latency sacrificing the storage utilization to reach the same accuracy. However, at high noise margin the Passive algorithm has the best reconstruction accuracy compared to other algorithms.
1 January 2026 in IEEE Access