Review of Computer Engineering Research

A data-driven model for student retention in a Philippine higher education institution

Bell S. Campanilla, Jonathan O. Etcuban, Patrick D. Cerna — Tue, 02 Dec 2025 00:00:00 -0600

The impact of higher education inevitably enhances economic, social, and human capital development. In the Philippines, many students require assistance to complete their college degrees, resulting in high dropout rates. The study aimed to determine the attributes that contributed to students' degree completion, predict them using the Decision Tree (DT) algorithm, and assist university policymakers with information to help them create effective early intervention plans for at-risk students. The Knowledge Discovery in Databases (KDD) process extracts accurate information from 3,417 student records, which consist of academic and socio-economic attributes, after undergoing selection, preprocessing, transformation, data mining, and interpretation/evaluation. MS Excel and RapidMiner were used to explore data, build predictive models, and generate insights. The results show that the DT model achieves an accuracy of 82.50%. To finish a degree, students must have a grade point average (GPA) of 2.55 or lower and be no older than 19. Parents' educational background and living outside the city also affect the decision outcomes. Living outside the city and parents' academic backgrounds also influence GPA outcomes. The researchers recommend reviewing, analyzing, and, as necessary, revising previous policies to enhance university programs that address this long-standing issue.

Blockchain-enabled secure EHR sharing with cloud storage using smart contracts and IPFS

Sunitha B J, Saravana Kumar S — Tue, 02 Dec 2025 00:00:00 -0600

The purpose of this study is to address the persistent security and privacy challenges in cloud-based Electronic Health Record (EHR) sharing by proposing a blockchain-enabled architecture that integrates decentralized storage and smart contracts. Traditional mobile cloud solutions improve data accessibility but rely on centralized control, making them vulnerable to unauthorized access, single points of failure, and limited patient transparency. To overcome these limitations, this research designs a user-centric access control framework that leverages the Ethereum blockchain, smart contracts, and the InterPlanetary File System (IPFS) within a mobile cloud environment. The methodology involves developing and deploying a prototype on Amazon Web Services, supported by an Android-based mobile application that enables healthcare providers and patients to interact with the blockchain network. Experimental evaluation was conducted using wearable sensor data to test the performance, scalability, and resilience of the proposed system. The findings indicate that the framework ensures secure EHR exchange, enforces fine-grained access policies, and achieves reduced latency compared to conventional centralized approaches. Unauthorized requests were reliably detected and blocked through the smart contract mechanism, while authorized users accessed records with minimal delay. The results also confirm the lightweight overhead of the system, making it practical for mobile healthcare environments. The practical implications of this work lie in offering a tamper-resistant, transparent, and patient-centric solution for medical data sharing, thereby improving trust, reducing administrative overhead, and supporting real-time healthcare services in distributed and resource-constrained settings.

A hybrid deep learning model for pneumonia detection in chest x-ray images

Namrata Ghuse, Rajkumar Jain, Sandeep Monga — Tue, 02 Dec 2025 00:00:00 -0600

Pneumonia is the leading respiratory cause of illness worldwide and has a significant impact on global health, challenging health systems, especially in resource-constrained scenarios, for immediate diagnosis. Timely and accurate diagnosis is essential to improve management and reduce mortality. Chest X-ray is a frequently used diagnostic tool, valued for its availability, rapidity, cost-effectiveness, and accuracy in detecting respiratory conditions. Recent developments in deep learning (DL) and machine learning (ML) have transformed the field of medical imaging, leading to improved diagnostics. In this context, a novel deep learning approach for automatic pneumonia detection in chest X-ray images is presented. Based on the advanced CNN architecture InceptionV3, this model is more effective at capturing features from data frames and employs LSTM networks to learn sequential data efficiently. This integration enables the model to learn both the spatial information of medical images and the temporal relationships, which enhances classification accuracy. Extensive experiments on public datasets with existing CNN-based models demonstrate that the proposed hybrid architecture surpasses traditional CNN models in accuracy, F1-score, and ROC-AUC-score, achieving 91.67%, 93.47%, and 97%, respectively. These results further support the potential of hybrid deep learning approaches as innovative methodologies for improving diagnostic accuracy and assisting healthcare professionals in lung infection diagnosis.

Iterative multi-scale deep learning framework for reliable breast cancer diagnosis

Komal S. Gandle, Dhananjay B. Kshirsagar — Tue, 02 Dec 2025 00:00:00 -0600

Accurate diagnosis of breast cancer from histopathological images is challenging due to variable tissue morphology and the subjectivity of manual interpretations. While CAD systems offer automated diagnosis, they often lack robust feature representation, contextual understanding, and integration of expert knowledge, limiting their effectiveness, especially in distinguishing invasive from pre-invasive carcinoma. This study presents a comprehensive deep learning-based diagnostic framework that integrates five novel modules to improve interpretability, feature robustness, and decision reliability. The Multiscale Attention Integrated Self-Supervised Representation (MAISSR) learns pathology-aware embeddings via co-optimized multiscale attention and contrastive learning. The Morphological-Geodesic Graph Convolutional Network (MG-GCN) combines geodesic topology with glandular morphology in a spatial graph model to capture epithelial transitions. The Hyper-Resolution Fusion Network with Cellular-Density Priors (HRF-CDPNet) enhances resolution in critical regions using cellular density maps. The Contextual Relational Transformer with Progression-Encoding (CRT-PE) models disease progression using spatial-contextual tokens to improve invasion mapping. Finally, Adaptive Cross-Modality Decision Calibration (ACM-DC) uses a reinforcement-learning-based agent to align machine predictions with expert annotations, especially in ambiguous cases. This integrated approach yields marked improvements in diagnostic metrics: F1 score increased from 82.5% to 89.3%, AUC from 0.88 to 0.94, and diagnostic agreement with experts from 85.2% to 94.8%. Overall, this work demonstrates the potential of a multi-factorial, multi-perspective framework to advance breast cancer diagnosis through optimized feature learning, spatial reasoning, and expert-machine synergy.

Machine learning from data to diagnosis: A comprehensive review of AI applications in mental health assessment

Rinki Kumari, Hitesh Marwaha — Tue, 02 Dec 2025 00:00:00 -0600

This systematic review examines machine learning and artificial intelligence applications in mental health disorder diagnosis, prediction, and management to identify current trends, knowledge gaps, and future research directions. We conducted comprehensive database searches across PubMed, IEEE Xplore, Scopus, and ScienceDirect from 2013-2024, following PRISMA guidelines for systematic review methodology. After screening 1,156 records, 46 peer-reviewed studies met inclusion criteria for analysis. Studies employed supervised learning methods, deep learning architectures, and natural language processing techniques across depression, anxiety, schizophrenia, bipolar disorder, autism spectrum disorder, ADHD, and OCD. Data sources included neuroimaging, wearable sensors, social media content, electronic health records, and multimodal data integration approaches. Depression and schizophrenia dominated research focus due to public health impact, with publication frequency increasing significantly after 2017. Most studies reported performance using accuracy, sensitivity, specificity, and F1-scores, though validation protocols varied considerably across investigations. Machine learning models demonstrated promising diagnostic accuracy and early detection capabilities across multiple mental health conditions. However, significant challenges persist including limited model generalizability, inconsistent external validation, data quality heterogeneity, algorithmic bias concerns, and clinical implementation barriers. Future research should prioritize developing explainable AI models, establishing standardized evaluation frameworks, implementing robust ethical guidelines, and fostering interdisciplinary collaboration between technology developers and healthcare providers to translate AI innovations into clinical practice for improved early detection, personalized treatment approaches, and enhanced diagnostic accuracy in mental healthcare.