Garuda - Garba Rujukan Digital

Diagnostic Accuracy of a Federated Learning Algorithm for Proliferative Diabetic Retinopathy Detection: A Multicenter Indonesian Study Rachmat Hidayat; Dedi Sucipto; Alexander Mulya; Ifah Shandy
Sriwijaya Journal of Ophthalmology Vol. 8 No. 2 (2025): Sriwijaya Journal of Ophthalmology
Publisher : Department of Opthalmology, Faculty of Medicine, Universitas Sriwijaya

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.37275/sjo.v8i2.136

Introduction: Proliferative diabetic retinopathy (PDR) remains a leading cause of preventable blindness in resource-limited settings. Federated learning (FL) enables collaborative artificial intelligence (AI) model training without sharing patient data. This study evaluated the diagnostic accuracy of an FL-based algorithm for PDR detection across three Indonesian ophthalmology centers. Methods: This multicenter, prospective, diagnostic accuracy study enrolled 512 eyes from 289 patients with type 2 diabetes at three private hospital ophthalmology clinics in Palembang, Jakarta, and Surabaya, Indonesia (January 2023–December 2024). All eyes underwent standardized fundus photography, spectral-domain optical coherence tomography, and comprehensive ophthalmic examination. The FL-based deep learning algorithm was evaluated against two independent retinal specialists using the International Clinical Diabetic Retinopathy classification. Primary outcomes were sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Results: The FL-based AI achieved a sensitivity of 94.6% (95% CI 81.8–99.3), specificity of 91.2% (95% CI 88.2–93.6), and AUC of 0.962 (95% CI 0.943–0.981) for PDR detection. Agreement with retinal specialists was substantial (κ = 0.87). Performance was consistent across centers (AUC 0.955–0.968; p = 0.841). Media opacity was the strongest predictor of misclassification (OR 3.42; 95% CI 1.87–6.25; p < 0.001). Conclusion: The FL-based AI demonstrated high diagnostic accuracy for PDR detection comparable to retinal specialists across multiple Indonesian centers. This privacy-preserving approach may facilitate scalable diabetic retinopathy screening in resource-limited ophthalmology settings.

Empowering Educators, Supporting Students: A Quasi-Experimental Evaluation of a Train-the-Trainer Model for School Mental Health in Indonesia Ahmad Badruddin; Omar Alieva; Ifah Shandy; Henny Kesuma; Benyamin Wongso; Winata Putri; Habiburrahman Said
Indonesian Community Empowerment Journal Vol. 5 No. 2 (2025): Indonesian Community Empowerment Journal
Publisher : HM Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.37275/icejournal.v5i2.49

Adolescent mental health is a pressing concern in urban Indonesian schools, where a significant gap exists between student needs and the availability of professional support. This study evaluated the efficacy of a culturally-adapted, school-based "Train-the-Trainer" (TtT) community service model designed to build sustainable mental health support capacity by empowering teachers. A quasi-experimental study with a matched control group was conducted in 20 public high schools in South Sumatera, Indonesia. Ten schools (n=150 teachers, n=1500 students) received the TtT intervention, where core teachers were trained to cascade mental health literacy and foundational support skills to their peers. Ten matched schools (n=145 teachers, n=1450 students) served as a control group. Data on teacher self-efficacy, student-reported support awareness, and school mental health policies were collected at baseline, 6-months, and 12-months. A linear mixed-effects model revealed a significant time-by-group interaction, with teachers in the intervention group reporting substantially higher confidence in supporting students at 12 months (M=4.15, 95% CI [4.01, 4.29]) compared to the control group (M=2.51, 95% CI [2.37, 2.65]), a large effect (d=2.41). Intervention students were significantly more likely to know how to access support (78% vs. 27%; OR=9.82, 95% CI [8.11, 11.89], p < 0.001). Intervention schools demonstrated a massive increase in formalized mental health protocols compared to control schools (IRR=7.94, p < 0.001). In conclusion, the TtT model is a highly effective and scalable strategy for building a foundational mental health support system within existing school structures in resource-constrained settings. By investing in local educators, this model fosters a sustainable, multi-tiered support ecosystem, offering a viable pathway for national policy and practice in Indonesia.

Neural Network Versus Stepwise Regression for Forensic Stature Estimation from Percutaneous Tibial Dimensions in South Sumatran Malay Adults Sari Sulistyoningsih; Abu Bakar; Eduardo Michael Perez; Ifah Shandy
Sriwijaya Journal of Forensic and Medicolegal Vol. 3 No. 2 (2025): Sriwijaya Journal of Forensic and Medicolegal
Publisher : Phlox Institute

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59345/sjfm.v3i2.248

Background. Stature estimation from skeletal elements underpins forensic biological profiling and is decisive in disaster victim identification, yet no validated osteometric standard exists for the South Sumatran Malay population, and stepwise regression assumes a linearity that skeletal growth biology does not obey. Methods. 450 healthy South Sumatran Malay adults (225 male, 225 female) aged 20-50 years were enrolled at CMHC Research Center, Palembang, and reported per STROBE. Five percutaneous right-tibial dimensions were measured by one certified anthropologist under the Martin and Saller convention (all relative technical errors of measurement below 1.5%), and a stratified 70:30 partition withheld 135 observations for testing. Stepwise multiple linear regression (SMLR) was compared against a grid-search-optimised 5-64-32-16-1 multilayer perceptron artificial neural network (MLP-ANN). Results. Sexual dimorphism was large for every dimension (Cohen's d 1.42-1.75; all p < 0.001). Percutaneous tibial length was the strongest single predictor (males r = 0.812, 95% CI 0.762-0.852). The calibration-corrected pooled regression reached R2 = 0.742 with RMSE ±4.82 cm, whereas the network reached R2 = 0.914 (95% CI 0.891-0.933) and RMSE ±2.78 cm — ΔR2 = 0.172 (bootstrap 95% CI 0.141-0.203), a 23.2% gain in explained variance and a 42.3% reduction in error that narrows the 95% forensic identification window from ±9.45 cm to ±5.45 cm. Conclusion. These first population-specific, artificial-intelligence-driven standards materially improve the biological profiling of incomplete human remains in Indonesian medicolegal and disaster victim identification practice.

Diagnostic Accuracy of a Federated Learning Algorithm for Proliferative Diabetic Retinopathy Detection: A Multicenter Indonesian Study Rachmat Hidayat; Dedi Sucipto; Alexander Mulya; Ifah Shandy
Sriwijaya Journal of Ophthalmology Vol. 8 No. 2 (2025): Sriwijaya Journal of Ophthalmology
Publisher : Department of Ophthalmology, Faculty of Medicine, Universitas Sriwijaya

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.37275/sjo.v8i2.136

Introduction: Proliferative diabetic retinopathy (PDR) remains a leading cause of preventable blindness in resource-limited settings. Federated learning (FL) enables collaborative artificial intelligence (AI) model training without sharing patient data. This study evaluated the diagnostic accuracy of an FL-based algorithm for PDR detection across three Indonesian ophthalmology centers. Methods: This multicenter, prospective, diagnostic accuracy study enrolled 512 eyes from 289 patients with type 2 diabetes at three private hospital ophthalmology clinics in Palembang, Jakarta, and Surabaya, Indonesia (January 2023–December 2024). All eyes underwent standardized fundus photography, spectral-domain optical coherence tomography, and comprehensive ophthalmic examination. The FL-based deep learning algorithm was evaluated against two independent retinal specialists using the International Clinical Diabetic Retinopathy classification. Primary outcomes were sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Results: The FL-based AI achieved a sensitivity of 94.6% (95% CI 81.8–99.3), specificity of 91.2% (95% CI 88.2–93.6), and AUC of 0.962 (95% CI 0.943–0.981) for PDR detection. Agreement with retinal specialists was substantial (κ = 0.87). Performance was consistent across centers (AUC 0.955–0.968; p = 0.841). Media opacity was the strongest predictor of misclassification (OR 3.42; 95% CI 1.87–6.25; p < 0.001). Conclusion: The FL-based AI demonstrated high diagnostic accuracy for PDR detection comparable to retinal specialists across multiple Indonesian centers. This privacy-preserving approach may facilitate scalable diabetic retinopathy screening in resource-limited ophthalmology settings.

Can Technical Indicators Predict Daily Price Direction? Walk-Forward Evidence from Bitcoin, the S&P 500, and Gold Ifah Shandy; Benyamin Wongso
Enigma in Economics Vol. 4 No. 1 (2026): Enigma in Economics
Publisher : Enigma Institute

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.61996/economy.v4i1.132

Background. Technical analysis remains among the most accessible forms of market decision support, but its incremental predictive value depends on design choices, and repeated model selection can create spurious backtest performance. Objective. This study evaluates whether widely used technical indicators, reconstructed from public formulas, can predict the next daily price direction of Bitcoin, the S&P 500, and gold. Methods. Daily data from 1 January 2015 to 14 July 2026 are examined using a chronological walk-forward design. Every signal observed at the close of day t is matched only with the sign of the close-to-close return from t to t+1. The main out-of-sample period begins in 2019, with 2019–2022 used for model selection and 2023–2026 reserved for confirmation. Performance is measured primarily by balanced accuracy and supplemented by accuracy, directional recall, stationary-bootstrap confidence intervals, and after-cost trading outcomes. Results. The best single indicator, Ichimoku 9/26/52, produced a macro balanced accuracy of 50.9%, while the best predetermined combination, Volume Confirmed, reached 50.8%. A new ridge-logistic hybrid indicator, SPAH-1, achieved 51.9% in validation and 51.3% in confirmation. Its confirmation balanced accuracy was 49.8% for Bitcoin, 49.2% for the S&P 500 proxy, and 55.0% for gold; only gold's bootstrap interval excluded 50%, but its predictions were strongly biased toward the upward class. After trading costs, Volume Confirmed underperformed buy-and-hold for all three assets. Additional selective experiments did not support an 80% daily prediction target. Conclusion. Technical indicators may assist regime description and decision confirmation, but they do not provide a robust universal next-day forecasting edge.

Title

Found 5 Documents
Search

Abstract

Abstract

Abstract

Abstract

Abstract

Title Search

Found 5 Documents Search

Abstract

Abstract

Abstract

Abstract

Abstract

Title

Found 5 Documents
Search