https://jeeemi.org/index.php/jeeemi/issue/feed Journal of Electronics, Electromedical Engineering, and Medical Informatics2026-11-18T06:11:02+07:00Dr. Triwiyantoeditorial.jeeemi@gmail.comOpen Journal Systems<p>The Journal of Electronics, Electromedical Engineering, and Medical Informatics, (JEEEMI), is a peer-reviewed periodical scientific journal aimed at publishing research results of the Journal focus areas. The Journal is published by the Department of Electromedical Engineering, Health Polytechnic of Surabaya, Ministry of Health, Indonesia. The role of the Journal is to facilitate contacts between research centers and the industry. The aspiration of the Editors is to publish high-quality scientific professional papers presenting works of significant scientific teams, experienced and well-established authors as well as postgraduate students and beginning researchers. All articles are subject to anonymous review processes by at least two independent expert reviewers prior to publishing on the International Journal of Electronics, Electromedical Engineering, and Medical Informatics website.</p>https://jeeemi.org/index.php/jeeemi/article/view/1998Neuro-Spectral Causal Fusion Network with Lesion Dependency Reasoning for Explainable Diabetic Retinopathy Detection2026-11-07T22:33:52+07:00K. Venuvenu.kalaimagal@gmail.comB. Thanga Parvathidrbtparvathi@gmail.comKunchanapalli Rama Krishnatenalirama@kluniversity.inYuvaraj S.yuvaraj.scse@sece.ac.in<p><strong>Background</strong>: Diabetic Retinopathy (DR) is a progressive microvascular complication of diabetes and a leading cause of preventable vision impairment worldwide. Early and accurate identification of DR severity is essential for timely intervention and effective disease management. However, conventional retinal screening relies heavily on manual examination of fundus images, which is labor-intensive, subjective, and dependent on specialist expertise. Although recent deep-learning models have demonstrated promising diagnostic performance, many existing approaches inadequately capture lesion-level relationships, underutilize clinical information, and provide limited interpretability for clinical decision-making.</p> <p><strong>Aims</strong>: this study proposes a novel Neuro-Spectral Causal Fusion Network (NSCF-Net) for explainable diabetic retinopathy detection and severity grading.</p> <p><strong>Methods</strong>: The proposed framework integrates an adaptive Neuro-Spectral Enhancement Module to improve lesion visibility, a Lesion Dependency Reasoning Graph to model pathological interactions among retinal abnormalities, and a Dual-Path Hybrid Transformer to jointly learn local lesion characteristics and global retinal context. Furthermore, a Causal Multimodal Fusion Unit combines retinal image representations with clinically relevant parameters, including HbA1c level, blood pressure, age, and duration of diabetes. To improve transparency and clinical trustworthiness, an Explainable Lesion Attribution Module generates counterfactual attention maps and hierarchical saliency visualizations highlighting diagnostically significant retinal regions. The proposed framework was evaluated on the APTOS 2019 diabetic retinopathy dataset, consisting of 3,662 retinal fundus images distributed across five DR severity classes.</p> <p><strong>Results</strong>: Experimental results demonstrate the effectiveness of NSCF-Net, achieving 99.1% accuracy, 98.9% precision, 99.0% recall, and 98.8% F1-score. Comparative analyses and ablation studies further confirm the contribution of neuro-spectral enhancement, lesion dependency reasoning, multimodal fusion, and explainability components to overall performance improvement.</p> <p><strong>Conclusion</strong>: These findings indicate that NSCF-Net provides an accurate, interpretable, and clinically reliable framework for automated diabetic retinopathy screening, severity assessment, and computer-aided ophthalmic decision support.</p>2026-11-04T00:00:00+07:00Copyright (c) 2026 K. Venu, B. Thanga Parvathi, Kunchanapalli Rama Krishna, Yuvaraj S.https://jeeemi.org/index.php/jeeemi/article/view/2012AEPS-PINN: Adaptive Energy-Prior Scheduling with Superposed Multi-Harmonic Excitation for Physics-Informed Electrical Impedance Tomography2026-11-18T06:11:02+07:00Dikky Juliyantodikky.23108@mhs.unesa.ac.idEndah Rahmawatiendahrahmawati@unesa.ac.id<p><strong>Background</strong>: Electrical Impedance Tomography (EIT) reconstructs internal conductivity distributions from boundary voltage measurements, but its ill-posed nature and limited sensitivity to interior structures remain major challenges.</p> <p><strong>Aims</strong>: This study develops and evaluates AEPS-PINN, a physics-informed reconstruction framework incorporating Adaptive Energy-Prior Scheduling to reduce the computational overhead of energy-based regularization.</p> <p><strong>Methods</strong>: The framework combines superposed multi-harmonic excitation, a dual-head Physics-Informed Neural Network (PINN) for electric potential and conductivity estimation, and an Energy-Based Model (EBM) prior activated selectively when the governing partial differential equation residual exhibits stagnation. Training employs Adam optimization followed by L-BFGS refinement.</p> <p><strong>Results</strong>: Performance was evaluated using a synthetic three-inclusion conductivity phantom under identical optimization budgets for Standard PINN, Static EBM-PINN, and AEPS-PINN configurations. AEPS-PINN achieved a mean squared error (MSE) of 0.9008 (n = 5), compared with 0.9127 (n = 4) for Standard PINN and 0.8967 (n = 5) for Static EBM-PINN, with no statistically significant improvement over Standard PINN under the Wilcoxon signed-rank test. The adaptive scheduler reduced EBM evaluations by approximately 99.8%, from 7,511 ± 353 to 15 ± 4 evaluations per run, without a measurable deterioration in reconstruction accuracy. However, Dice and Intersection-over-Union scores remained zero across all configurations, indicating unsuccessful spatial recovery of the conductivity inclusions. Furthermore, the superposed multi-harmonic excitation produced a rank-1 boundary-voltage response matrix, corresponding to a single independent current-injection pattern rather than four separately applied patterns.</p> <p><strong>Conclusion</strong>: These findings demonstrate that AEPS-PINN substantially reduces energy-prior evaluation requirements while maintaining reconstruction performance comparable to continuously regularized PINN. Nevertheless, reliable inclusion recovery requires more informative boundary measurements, particularly multiple independent current-injection patterns, followed by experimental validation.</p>2026-11-11T00:00:00+07:00Copyright (c) 2026 Dikky Juliyanto, Endah Rahmawatihttps://jeeemi.org/index.php/jeeemi/article/view/1921Comparative Cyclic Mechanical Response of Commercial SACH Prosthetic Feet under ISO 10328-Based Fatigue Loading2026-11-17T09:17:20+07:00Mahfud Saifudinsaepsyaifudin@gmail.comNur Rachmatnurrachmat@gmail.comSisybania Sisybaniasisybania@gmail.comAnissa Eka Septianisanissaeka@gmail.com<p><strong>Background</strong>: The Solid Ankle Cushion Heel (SACH) foot remains widely prescribed in lower-limb prosthetic rehabilitation, particularly in low- and middle-income countries, because of its simple construction, low maintenance demand, and affordability. However, commercially supplied SACH feet that appear externally similar may respond differently to repetitive loading, and procurement decisions are frequently based on internal inventory labels rather than independent mechanical evidence. Previous studies have mainly characterized prototype designs or single reference feet under static or quasi-static loading, so comparative cyclic evidence for finished commercial SACH feet remains scarce.</p> <p><strong>Aims</strong>: This exploratory study aimed to compare the early observable cyclic mechanical response of two commercial SACH foot groups and to establish a reproducible screening framework for finished commercial components.</p> <p><strong>Methods</strong>: Six new right-sided SACH feet were tested: three Group B specimens (sizes 23R, 24R, and 25R) and three size-matched Group A specimens. A pneumatic fatigue tester applied alternating heel and forefoot loading following the principles of ISO 10328:2016, Section 17.2.5, with a target range of 50–1,330 N at 1 Hz; the recorded maximum load was 1,460 N for all specimens, representing a 9.8% overshoot. The predefined endpoint was the first visible microdeformation, and specimens without an event were treated as right-censored observations.</p> <p><strong>Results</strong>: All three Group B specimens (3/3, 100%) developed visible dorsal microdeformation at 10,000 cycles, whereas none of the Group A specimens (0/3, 0%) showed a comparable abnormality through 100,000 cycles. One Group B specimen was inadvertently continued to 150,000 cycles after its endpoint because of a protocol communication error. No specimen developed cracking, delamination, permanent deformation, loss of fit, or structural failure.</p> <p><strong>Conclusion</strong>: The scientific contribution of this study is a comparative, endpoint-driven evaluation of finished commercial SACH feet, rather than prototype development, under identical cyclic loading. The findings indicate a consistent difference in early cyclic response under the applied protocol. However, the small sample, visual endpoint assessment, load overshoot, unequal test exposure, and absence of testing to failure preclude conclusions regarding comparative fatigue life or ISO conformity.</p>2026-11-17T09:12:42+07:00Copyright (c) 2026 Mahfud Saifudin, Nur Rachmat, Sisybania Sisybania, Anissa Eka Septianihttps://jeeemi.org/index.php/jeeemi/article/view/2105An Explainable Deep Learning Genomic Framework for Dengue Virus Serotype Classification and Phylogenetic Analysis2026-11-18T06:01:43+07:00Sony Ahujaahujasony.cse@gmail.comDeepti Shrimankardshrimankar@cse.vnit.ac.inAditi Durgedurge.cse@gmail.com<p><strong>Background</strong>: Accurate dengue virus genomic surveillance requires reliable serotype identification together with evolutionary and interpretable analysis. However, existing approaches often address classification, phylogenetic analysis, or explainability independently and may not provide an integrated framework capable of jointly handling sequencing artifacts, genomic representation learning, serotype classification, evolutionary inference, and model interpretation.</p> <p><strong>Aim</strong>: Therefore, this study aims to develop a Deep Learning Hybrid Genomic Framework (DLHGF) that integrates genomic preprocessing, Transformer-based representation learning, graph-based serotype classification, phylogenetic inference, and explainable artificial intelligence within a unified analytical workflow for dengue virus genomic analysis.</p> <p><strong>Methods</strong>: The framework applies a Generative Adversarial Network-based Deep Genome Cleaner for sequence denoising, followed by MAFFT-L-INS-i multiple sequence alignment. A Genomic Transformer Encoder learns representations from overlapping 3-mer tokens. The resulting embeddings are processed using a Hierarchical Graph Neural Network integrated with a one-dimensional Convolutional Neural Network for serotype classification. An Adaptive Phylogenetic Maximum Likelihood module performs evolutionary analysis using aligned sequences, while Gradient-weighted Class Activation Mapping++ identifies genomic regions influencing classification. The framework was evaluated using 10,000 complete DENV genomes, with 2,500 sequences per serotype.</p> <p><strong>Results</strong>: DLHGF achieved 98.5% classification accuracy and an F1-score of 0.95. The phylogenetic module achieved a mean raw Robinson–Foulds distance of 0.92. The complete framework required approximately 2.4 hours for training and 0.18 seconds per sequence for inference on an NVIDIA A100 GPU.</p> <p><strong>Conclusion</strong>: DLHGF integrates serotype classification, phylogenetic inference, and model interpretability within a single dengue genomic workflow. The random-split results represent an in-distribution benchmark, while broader lineage-independent, geographic, temporal, and prospective validation remains necessary to assess generalizability.</p>2026-11-18T05:39:21+07:00Copyright (c) 2026 Sony Ahuja, Deepti Shrimankar, Aditi Durge