Sriwijaya Journal of Radiology and Imaging Research

Diffusion Tensor Imaging and Resting-State Functional MRI Reveal Coupled Microstructural and Default-Mode Network Alterations in Mild Cognitive Impairment: A Diagnostic Accuracy Study

Taryudi Suharyana — 2026-07-09

Introduction: Early-stage neurodegeneration — clinically expressed as mild cognitive impairment (MCI) — lacks accessible imaging biomarkers that link microstructural white-matter injury to functional network disruption. We evaluated whether integrating diffusion tensor imaging (DTI) and resting-state functional MRI (rs-fMRI) at 3.0-Tesla improves detection of MCI and substantiates a structural-to-functional disconnection mechanism.

Methods: In this STARD-2015-compliant, prospective cross-sectional diagnostic accuracy study at a tertiary hospital in Palembang, Indonesia, 85 participants (45 MCI, 40 age-, sex- and education-matched controls) underwent single-scanner 3.0-T MRI. DTI metrics (fractional anisotropy [FA], mean diffusivity [MD]) were derived by tract-based spatial statistics and posterior-cingulate-seeded Default Mode Network (DMN) connectivity by CONN. Consensus clinical diagnosis (Petersen criteria) was the blinded reference standard. Diagnostic accuracy used ROC/DeLong AUC, Wilson 95% confidence intervals (CIs), Cohen κ, and multivariable logistic regression.

Results: Posterior-cingulum FA was lower (0.399 ± 0.058 vs 0.489 ± 0.052, p<0.001) and MD higher in MCI; PCC–mPFC connectivity was reduced (0.392 ± 0.124 vs 0.586 ± 0.136, p<0.001). FA discriminated MCI with AUC 0.903 (95% CI 0.831–0.974; sensitivity 82.2%, specificity 90.0%); a combined FA+FC model reached AUC 0.901 with sensitivity 95.6% and NPV 93.8% but did not exceed FA alone (DeLong p=0.90). FA and connectivity were strongly correlated (r=0.79, 95% CI 0.69–0.86, p<0.001). Inter-reader agreement was substantial (κ=0.74 and 0.67).

Conclusion: Multimodal 3.0-T DTI and rs-fMRI provides an accurate, radiation-free signature of early-stage neurodegeneration; the coupling between cingulum microstructure and DMN connectivity is consistent with structural disconnection being associated with functional decoupling and offers a deployable tool for tertiary referral centres.

Deep Learning-Based Image Enhancement of Low-Field Brain MRI: A Multicenter Validation of Diagnostic Image Quality and Lesion Detection

Rachmat Hidayat — 2026-07-09

Introduction: Low-field MRI (LF-MRI) widens neuroimaging access in resource-limited settings but suffers low signal-to-noise ratio (SNR), reduced resolution and artefacts. We developed and validated a deep-learning framework for image normalisation and noise reduction to elevate 0.35T brain MRI toward high-field quality, and tested whether it improves clinically significant lesion detection.

Methods: In a multicentre retrospective diagnostic-accuracy study (STARD 2015), 450 adults underwent non-contrast 0.35T brain MRI (T1W, T2W, FLAIR) across three private tertiary centres in Palembang, Indonesia. Images were enhanced with a CycleGAN incorporating Vision-Transformer blocks. Three blinded neuroradiologists scored a 5-point Likert scale and recorded lesion presence; paired 1.5T MRI was the reference standard. Sensitivity, specificity, AUC and likelihood ratios were computed with 95% CIs; tests compared by McNemar and DeLong; agreement by Fleiss kappa.

Results: AI enhancement improved all quality metrics (e.g., T1W PSNR 22.15 to 28.45 dB; SSIM 0.71 to 0.89; all p<0.001). For lesion detection, AI-enhanced LF-MRI achieved sensitivity 93.9% (95% CI 89.4–96.6), specificity 91.1% (87.1–94.0) and AUC 0.94 (0.91–0.97) versus 78.3%, 81.1% and 0.81 for original images (DeLong p<0.001; McNemar p<0.001). LR+ rose to 10.56 and LR− fell to 0.067. Inter-reader agreement was almost perfect (Fleiss kappa 0.78–0.85).

Conclusion: A CycleGAN-with-transformer framework substantially improved objective quality and diagnostic performance of 0.35T brain MRI toward high-field standards with almost-perfect reader agreement. Pending prospective and external validation, AI enhancement is a low-cost route to more equitable neuroimaging.

Deep-Learning Pharmacokinetic Modelling for Personalised Low-Dose Gadolinium in Oncological 3.0-T MRI: A Diagnostic-Accuracy Study

Oliva Azalia Putri — 2026-07-13

Introduction: Weight-based dosing of gadolinium-based contrast agents (GBCA; 0.1 mmol/kg) in oncological MRI disregards individual haemodynamics and tumour microvascularity, contributing to avoidable cumulative exposure and tissue-retention risk. We evaluated an artificial-intelligence (AI)-assisted pharmacokinetic-modelling approach to personalise and reduce GBCA dose while preserving diagnostic performance.

Methods: In this prospective, paired diagnostic-accuracy study reported per STARD 2015 at a tertiary hospital in Palembang, Indonesia, 152 adults (198 lesions) with histologically confirmed solid primary malignancies underwent 3.0-T contrast-enhanced MRI. A convolutional-neural-network extended-Tofts model derived each patient’s minimum effective gadobutrol dose, compared against the standard 0.1 mmol/kg protocol. Two radiologists, blinded to clinical data and to a composite reference standard (histopathology and ≥6-month imaging follow-up), assessed signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), 5-point diagnostic confidence and lesion characterisation. Sensitivity, specificity, AUC (DeLong), likelihood ratios, Cohen’s κ and McNemar testing were computed with 95% confidence intervals.

Results: The AI protocol reduced GBCA dose by 38.4% (4.6 vs 7.5 mL; p<0.001). Sensitivity was 95.2% (95% CI 90.9–97.6), specificity 83.3% (66.4–92.7) and AUC 0.94 (0.90–0.97) versus 0.95 (0.91–0.98) for standard dose (DeLong p=0.620; McNemar p=0.773). SNR and CNR were non-inferior (all p>0.05). Inter-reader agreement was substantial-to-almost-perfect (characterisation κ 0.83; confidence κ 0.88). Diagnostic adequacy was maintained in 149/152 cases (98%).

Conclusion: AI-assisted pharmacokinetic modelling enabled a 38% gadolinium-dose reduction without loss of diagnostic accuracy or image quality, supporting personalised contrast administration and lower cumulative exposure in oncological MRI.

Radiomics-Based Machine Learning for Automated Detection and Rupture-Risk Stratification of Cerebral Vascular Malformations: A Retrospective Cohort Study

Hesti Putri — 2026-07-14

Introduction: Cerebral vascular malformations (CVMs) are the leading cause of spontaneous intracranial haemorrhage, yet their detection on CT/MR angiography is operator-dependent and existing machine-learning models derive almost exclusively from Caucasian or East Asian cohorts. We developed and internally validated a population-specific radiomics pipeline for CVM detection and rupture-risk stratification in a Southeast Asian population.

Methods: In this retrospective diagnostic-and-prognostic cohort (STARD 2015; CLAIM) at a tertiary hospital in Palembang, Indonesia (2020–2025), 486 adults with diagnostic-quality CTA or TOF-MRA were analysed. After resampling and normalisation, 1,218 PyRadiomics features were reduced by LASSO and used to train Random Forest, SVM and XGBoost models (70/30 split). The reference standard was blinded consensus segmentation by two consultant neuroradiologists. Diagnostic accuracy (Wilson CI), AUC (DeLong), likelihood ratios, Cohen κ, McNemar test and a multivariable rupture model were computed.

Results: CVM prevalence was 42.8% (208/486). XGBoost was the best detector (AUC 0.963 (95% CI 0.950–0.977); sensitivity 88.5 (95% CI 83.4–92.1)%; specificity 92.4 (95% CI 88.7–95.0)%; LR+ 11.71 (95% CI 7.74–17.72)), outperforming a single radiologist (ΔAUC 0.128, p<0.001; McNemar χ²=9.72, p=0.002 (discordant pairs b=73, c=39)). Inter-reader agreement was almost perfect (κ 0.845 (95% CI 0.797–0.893)). A radiomics-clinical model stratified rupture (AUC 0.846 (95% CI 0.792–0.901)), with lesion size (OR 4.26 (95% CI 2.60–6.98)) and hypertension (OR 2.46 (95% CI 1.18–5.14)) dominant and good calibration (χ²=14.87, df=8, p=0.062).

Conclusion: A population-specific radiomics machine-learning pipeline achieved high diagnostic accuracy for CVM detection and clinically useful rupture-risk stratification, supporting operator-independent neurovascular triage in Southeast Asian settings. External validation is warranted.

Longitudinal MRI-PDFF Quantification of Hepatic Steatosis Reversibility During SGLT2-Inhibitor versus GLP-1 Receptor Agonist Therapy: A Prospective Cohort Study

Arsan Saliha — 2026-07-15

Introduction: Precise quantification of hepatic steatosis is central to managing metabolic dysfunction-associated steatotic liver disease (MASLD). MRI proton density fat fraction (MRI-PDFF) has displaced biopsy as a reproducible biomarker, yet real-world comparative monitoring data and formal early-response prediction remain scarce in Indonesian practice. We quantified steatosis reversibility on serial MRI-PDFF during SGLT2-inhibitor versus GLP-1 receptor agonist therapy and tested whether an early scan predicts response.

Methods: In this prospective cohort at a tertiary hospital in Palembang, Indonesia, 90 adults with MASLD (baseline PDFF ≥5.5%) were grouped by prescribed therapy (SGLT2i, n=45; GLP-1 RA, n=45) and imaged at baseline, Month 3 and Month 6 on a 3.0 T scanner using a confounder-corrected 3D multi-echo spoiled gradient-echo sequence. Two blinded radiologists measured PDFF across Couinaud segments V/VI/VIII; response was a ≥30% relative reduction. Analyses (STARD 2015) included linear mixed-effects modelling, κ and intraclass correlation, ROC (DeLong), likelihood ratios and multivariable logistic regression.

Results: PDFF fell in both arms (both p<0.001; partial η²=0.66 and 0.79). Responder rate was 75.6% (95% CI 61.3–85.8) with GLP-1 RA versus 40.0% (27.0–54.5) with SGLT2i (OR 4.64; p=0.001). A Month-3 decline ≥16.9% predicted Month-6 response with AUC 0.895 (0.823–0.967), sensitivity 96.2%, specificity 78.9%, LR+ 4.57, LR− 0.05. Inter-reader agreement was near-perfect (ICC 0.986; κ 0.861). GLP-1 RA therapy and higher baseline PDFF independently predicted response.

Conclusion: Serial MRI-PDFF reliably quantified pharmacologically-induced steatosis reversibility, with GLP-1 RA producing greater fat loss and an early scan accurately triaging responders. MRI-PDFF is a robust, decision-useful monitoring biomarker deployable in tertiary settings.