This paper addresses the limitations of existing deep learning techniques in retinal disease diagnosis from fundus images, specifically focusing on inadequate feature representation, localization challenges, and the lack of interpretability. Current methods also struggle with class imbalance and the need for simultaneous multi-task learning. To overcome these issues, we propose a novel interpretable deep learning framework for the simultaneous segmentation of multiple retinal lesions and multi-categorical disease classification with grading. Our framework incorporates a modified Convolutional Neural Network (CNN) with separable convolution layers for efficient parameter utilization and an attention mechanism to focus on critical image regions, enabling accurate lesion mapping. To tackle class imbalance, the framework utilizes the Synthetic Minority Over-Sampling Technique (SMOTE). Furthermore, Explainable AI (XAI) techniques, including Grad-Cam, are integrated to enhance the model's interpretability and provide insights into its decision-making process. The effectiveness of the proposed framework is evaluated on several publicly available datasets, including ODIR, RFMiD, and IDRiD. The framework aims to provide a comprehensive, balanced, and interpretable solution for real-time fundus disease diagnosis in healthcare. The proposed architecture outperforms existing models in accuracy, AUC, mAUPR, and mIoU on retinal fundus image datasets. Heatmaps and segmentation masks confirm their effectiveness in identifying disease regions and segmenting lesions.

Coral reefs have emerged as the most biodiverse and important entities in the marine ecosystem, as they house 25% of all marine organisms. As water temperatures rise in some sea areas, coral reef colors gradually turn white. This phenomenon, known as coral bleaching, signifies the deterioration of coral reef health and poses a significant threat to their survival. There is an urgent need for rapid and effective solutions to mitigate these threats, limit the spread of bleaching, and protect coral reefs. This study proposes a novel system that utilizes deep learning and computer vision to assess coral reef health and detect early signs of bleaching. Focusing on coral reefs in the Red Sea, the YOLOv8 and YOLOv9 object detection models were used on an augmented dataset of 10,285 labeled images representing healthy, bleached, and dead corals. The system includes a user-friendly interface for image classification and automatic notification of relevant authorities upon detection of bleaching or death. The evaluation results showed that YOLOv9 achieved a slightly higher mean Average Precision (mAP) of 89% compared to YOLOv8 (88%), demonstrating the effectiveness and potential of the system for real-time coral reef monitoring. This research offers a practical, automated solution for early detection, reducing human effort and achieving faster results, ultimately saving coral reefs from irreversible damage.

Effective requirements prioritization is a success factor for software development. Models of prioritization that are traditionally used like MoSCoW, RICE, AHP and the Kano Model are deficient in responding to what is essential for startups in terms of speed and delivery of value, efficiency in cost, and mitigation of risk. In this paper, we introduce RECTV, a new startup requirements prioritization framework. Essential decision making components including Cost, Risk, Technical Feasibility, Confidence Level (C), and Time to Value (TtV) are integrated as key pieces of the RECTV tool to help startups prioritize based on informed, data based decision making. In contrast to traditional models, RECTV validates the outcomes of the features while delivering business value in a way where startups can adapt to market feedback and uncertainty rapidly. This work positions RECTV as a practical alternative to traditional models, offering startups a structured and responsive approach to prioritize features under real-world constraints. By conducting a comparative analysis with other models, the framework is evaluated and proved to be superior at tackling the startup glitches such as low funds, unspecified features, and the necessity of fast market adjustment. For startups, RECTV helps them to make better use of resources and develop features that avoid unnecessary risks.

Early and accurate cancer detection is crucial for effective treatment, prognosis, and the advancement of precision medicine. Analyzing omics data is vital in cancer research. While using a single type of omics data provides a limited perspective, integrating multiple omics modalities allows for a more comprehensive understanding of cancer. Current deep models struggle to achieve efficient dimensionality reduction while preserving global information and integrating multi-omics data. This often results in feature redundancy or information loss, overlooking the synergies among different modalities. This paper proposes a contrastive adversarial encoder (CAEncoder) for multi-omics data integration to address this challenge. The proposed model combines a Vision Transformer (ViT) and a CycleGAN, trained in an end-to-end contrastive manner. The ViT is the encoder, utilizing self-attention, while the CycleGAN employs adversarial learning to ensure more discriminative and invariant latent space embeddings. Contrastive adversarial training improves representation quality by preventing information loss, eliminating redundancy, and capturing the synergies among different omics modalities. To ensure contrastive adversarial training, a composite loss function is used, consisting of a weighted combination of Adversarial Loss (Hinge Loss), Cycle Consistency Loss, and Triplet Margin Loss. The Adversarial Loss and Cycle Consistency Loss provide feedback from the CycleGAN, ensuring effective adversarial learning. Meanwhile, the Triplet Margin Loss promotes contrastive learning by pulling similar samples together and pushing dissimilar samples apart in the latent space. The performance of the CAEncoder is evaluated on downstream classification tasks, including both binary and multi-class classifications of five different cancer types. The results show that the model achieved a classification accuracy of up to 93.33% and an F1 score of 92.81%, outperforming existing advanced models. These findings demonstrate the potential of our method to enhance precision medicine for cancer through improved multi-omics data integration.

Software fault localization is critical for reducing debugging effort, still much of the current literature focuses on a single major source of evidence and limited to fine-grained localization scenarios. Spectrum-based techniques are very efficient in using test execution behavior, but they can often do poorly when failures propagate across structurally related statements or when lexical clues provided by bugs are present in both text and source-code artifacts related to a bug. To overcome this limitation, this research presents RLSFLoc, which is a reinforcement learning based semantic structural framework to fault localization at the statement level. The proposed framework is a combination of three evidence streams: execution-based suspiciousness through passing and failing tests, structural dependencies refinement based on program relations, and semantic relevance computed based on bug-related textual artifacts and program entities. Instead of using fixed weights to combine these signals, RLSFLoc uses a Proximal Policy Optimization-based adaptive fusion policy to learn context-sensitive weighting in various localization scenarios. The framework was evaluated on Defects4J v3.0.1 with 854 active bugs and statement-level localization setting. The experimental results indicated that RLSFLoc had a high performance compared to the compared methods with 48.7% Top-1 accuracy, an MRR of 0.612, a MAP of 0.587, and the lowest EXAM score of 0.184. Also, proposed approach minimized the inspection workload and ensured the performance (which was consistent across various projects) was not affected. Overall, the results indicate that a successful and practical approach to localizing software faults on a fine-grained level is to adaptively integrate the execution-based, structural and semantic evidence.