In this investigation, a Deep Learning (DL) approach was applied to measure the morphometry of Transverse Aeolian Ridges (TARs) on the surface of Mars. A large sample of TARs was segmented from High Resolution Imaging Science Experiment (HiRISE) images, the highest resolution remote sensing dataset presently available for the planet. HiRISE images located between 50°N and 50°S, and from all longitudes were selected. Morphometric parameters such as area, elongation, and orientation were retrieved for this sample using a supervised instance segmentation and geospatial analysis pipeline. The result is the most extensive catalogue of TAR morphometry to date extracted from ~14 million candidate TARs in ~7000 HiRISE Images. This was accomplished by training off-the-shelf DL models within a Geographic Information System (GIS) environment. A significant TAR population was found in approximately half of the images surveyed. TAR area, and the lengths of the long and short axes, were found to exhibit a positively skewed log-normal distribution; the median short axis length is 5 m, while the median long axis is 24 m. Median elongation is 0.24. Global TAR orientations are varied, although North-South oriented TARs are the most populous group. This is likely due to the strong east blowing winds predicted by GCM simulations of the modern martian climate. Here we present our latest results and use TAR orientation statistics to describe the emerging picture of global wind patterns on Mars during TAR forming epochs. • TARs have been segmented with a Mask R-CNN model, using ArcGIS Pro. • 14 million TARs were measured in ~7000 HiRISE images of the mid latitudes of Mars. • Orientation and morphometry data were retrieved using an automated DL/GIS pipeline. • TAR orientation provides a geomorphic markers for past wind conditions on Mars. • Comparisons to manual digitisations show the results to be representative and fit for purpose.

Deep Learning-Driven Transmission Electron Microscopy Analysis of Murine Optic Nerve Myelinated Axons.

• Non-destructive microfluidic imaging for pore-scale tracking of microplastics. • Integrated visualization–quantification framework links pore morphology to MP fate. • Heterogeneous pore networks act as long-term sinks for microplastics. • Connectivity-driven networks act as high-mobility conduits for microplastics. • Uniform throat networks act as reversible reservoirs of microplastic retention. Microplastics are persistent contaminants in soils and aquifers, accumulating and potentially remobilizing within groundwater and surface flows. Predicting their fate requires pore-scale insights in realistic porous media, especially under saturation-desaturation conditions. Despite the growing body of research on MP transport in porous media, the methodological divide between qualitative visualization and quantitative measurement limit predictive understanding under realistic subsurface conditions. Microfluidic systems enable direct pore-scale observation of MP mobility but rarely provide domain-wide quantification, while column-scale experiments offer robust mass balance yet obscure the underlying pore-scale mechanisms. This disconnect hinders the development of mechanistic understanding linking pore geometry, multiphase flow, and MPs fate. To address this limitation, this study introduces an integrated, non-destructive microfluidic imaging framework that couples high-resolution time-lapse visualization with semi-automated image analysis to simultaneously resolve pore-scale processes and quantify domain-scale retention across saturation-desaturation cycle. Four statistically distinct micromodels representing heterogeneous porous domains were used to capture realistic flow and retention behaviors. The segmentation pipeline classified individual phases (air, water, solid, and MPs), allowing spatially and temporally resolved quantification of retention and release throughout the entire flow field. The results demonstrate that MPs mobility depend not on porosity alone but on the combined effects of throat-size distribution, connectivity, and capillary dynamics, which control the persistence or remobilization of particles under transient saturation. Beyond its immediate findings, this work establishes a reproducible and scalable experimental-computational framework that bridges the gap between visualization and quantification, offering a mechanistic basis for predicting MPs fate in soils and aquifers under dynamic environmental conditions.

Ischemic stroke represents a leading cause of global mortality and long-term disability. Consequently, rapid quantification of cerebral damage severity, coupled with timely neuromodulatory intervention, is imperative for improving clinical prognosis. In this study, we present a label-free quantitative monitoring framework utilizing optical-resolution photoacoustic microscopy (OR-PAM) to assess microvascular alterations and evaluate the therapeutic efficacy of low-intensity transcranial ultrasound stimulation (LITUS). A graded ischemic stroke model was established by modulating photothrombotic irradiation duration (3, 5, and 10 min) to validate system sensitivity. Subsequently, a Hessian filter-based segmentation pipeline was employed to extract quantitative vascular metrics. Five key morphological and functional parameters, including the photoacoustic (PA) signal intensity, vessel area fraction (VAF), average vessel diameter (AVD), branch-point number (BN), and perfused vessel density (PVD), were extracted to characterize the severity of ischemic injury. Our analysis revealed a strong negative correlation between irradiation duration and vascular perfusion-related metrics, which was further confirmed by ex vivo 2,3,5-triphenyltetrazolium chloride (TTC) staining showing duration-dependent infarct expansion. Applying this quantitative framework to the established model, we further demonstrated that acute LITUS treatment significantly alleviated microvascular hypoperfusion and promoted rapid hemodynamic recovery by restoring both functional perfusion and vascular morphology via vasodilation. These findings highlight the potential of a PAM-based quantitative framework for accurate grading of ischemic severity and evaluation of ultrasound-based neuromodulation.

Deploying conditional Generative Adversarial Networks (cGANs) for industrial texture synthesis faces two barriers: the prohibitive cost of manual data annotation and the uncertain alignment between automated evaluation metrics and human perception. This study addresses both challenges for marble texture synthesis using 289 high-resolution industrial scans. We adapt an unsupervised segmentation pipeline combining Simple Linear Iterative Clustering (SLIC) superpixels, Gaussian Mixture Models (GMMs), and graph cut optimization to extract vein structures without manual annotation. Four cGAN architectures—baseline cGAN, Pix2Pix, BicycleGAN, and GauGAN—are benchmarked using a dual-evaluation protocol contrasting ten automated metrics with structured human-centered assessment. The results reveal a significant metric–perception discrepancy. Pix2Pix achieved the best Fréchet Inception Distance (FID = 85.3) yet received the lowest human ratings due to periodic texture artifacts. GauGAN produced textures statistically indistinguishable from real marble, achieving a Visual Turing Pass Rate (VTPR) of 0.533 and a Mean Opinion Score on Marble Authenticity (MOS-MA) of 2.89, despite an inferior FID (87.3). These findings make three contributions: an annotation-free segmentation pipeline, empirical evidence that automated metrics alone are insufficient for architecture selection, and a dual-evaluation framework that establishes human-in-the-loop assessment as essential for quality-critical industrial deployment.

Under growing urbanization and environmental challenges, sustainable urban mobility has become a critical priority for cities worldwide. Public Transport (PT) systems play a central role in reducing car dependency, lowering emissions, increasing network capacity, and promoting more equitable and efficient access to urban spaces for all users. Hence, the present paper aims to investigate PT preferences in the city of Larissa, Greece. Larissa is a medium-sized city currently serviced only by buses, and is currently focusing on the potential introduction of a new tram system to operate in parallel with existing bus services. To this end, a SP survey was designed and implemented, resulting in 972 observations that were collected for further statistical analysis. Survey results show a slight preference for trams over buses, with 54.63% selecting the tram and 45.37% favoring the buses. Moreover, a context-based segmentation pipeline was established using PCA, DBSCAN and t-SNE algorithms, aiding the visualization of existing clusters for transport choice approaches. Afterwards, a series of mixed logit models was applied, and statistically significant variables influencing mode choice were determined. The study also examines Value of Time (VoT) metrics and finds that respondents assign lower VoTs to trams than to buses, especially in out-of-vehicle segments of the journey, such as waiting and walking, and therefore consider trams as more pleasant and less burdensome. The findings also indicate that passengers place a high value on the quality of infrastructure related to access and waiting times, underlining the need to improve the overall user experience beyond the vehicle itself. In summary, the present research offers valuable insights into how the introduction of a tram system could possibly reshape PT usage patterns when compared with the legacy existing bus services.

ABSTRACT Extensive resin wood formation in Scots pine stems infected by Cronartium pini constitutes a significant quality defect that compromises timber value and complicates industrial processing decisions. While X-ray computed tomography (CT) offers non-destructive density mapping capabilities, the wood processing industry currently lacks validated, automated segmentation methods to differentiate resin wood from unaffected heartwood and sapwood. A hybrid automated segmentation pipeline incorporating an adaptive per-slice Gaussian mixture model (GMM) was developed to combine statistical parameterisation of tissue-specific intensity distributions with multi-directional ray-casting for structural likelihood scoring. The framework utilises unsupervised image processing methodologies and morphological refinement to improve boundary delineation between complex internal wood features. Validation against manual annotations by trained technicians demonstrated high precision (0.94) for sapwood segmentation and substantial inter-rater agreement (Cohen’s κ = 0.819). However, precision for resin wood was limited to 0.27, reflecting the inherent difficulty of distinguishing pathological resin saturation from natural impregnation of heartwood. By implementing heatmap-based logic to overcome the limitations of simple intensity thresholding, this research establishes a foundational open-source pathway for automated internal resin wood mapping. The proposed pipeline offers a computationally efficient approach for research-scale analysis and provides a viable strategy towards potential value recovery in sawmills.

Diabetic foot ulcers, resulting from neuropathic and/or vascular complications in patients with diabetes mellitus, pose a major global health challenge. Early detection and consistent monitoring of wound progression are essential for timely intervention, effective treatment, and the prevention of severe complications such as amputation. In modern diabetic foot care, images captured using digital cameras and mobile phones are increasingly employed for remote wound assessment. In this context, automated segmentation of these wounds from such images plays a vital role by enabling objective and quantitative evaluation of wound areas-crucial for tracking the progression of healing over time. Recent years have witnessed growing interest in deep learning-based wound segmentation techniques, with a particular focus on models that are both computationally efficient and suitable for deployment on resource-constrained devices, including smartphones and point-of-care platforms. In this study, we propose a lightweight convolutional neural network (CNN) for diabetic foot wound segmentation that augments the U-Net architecture with ghost feature generation and Convolutional Block Attention Modules (CBAM) to improve computational efficiency and feature representation. The model was evaluated on a privately annotated dataset of 3450 diabetic foot wound images and compared against state-of-the-art architectures, including SegNet, U-Net, MobileNetV2, Mask R-CNN, and the domain-specific approach of Wang et al. We further investigated a fully automated two-step pipeline for wound segmentation incorporating a prior foot segmentation-based ROI detection. Using ROI detection, the proposed CNN achieved a precision of 85.13%, recall of 91.84%, Dice coefficient of 86.95%, and IoU of 77.23%. These results demonstrate competitive performance relative to high-capacity models while maintaining substantially reduced computational complexity, highlighting its suitability for real-time clinical deployment in low-resource environments.

Skin segmentation from clinical photography is a crucial step in dermatological image analysis. However, the variability in skin tones, lighting conditions, anatomical regions, and the presence of additional objects introduces significant challenges. Due to these complexities, the segmentation process is often performed manually, as developing an algorithm capable of handling such diverse conditions is particularly difficult. Recently, open-world foundation models have emerged, offering the potential to generalize across diverse and unseen conditions. These models present a promising opportunity for dermatology. In this work, we adopt two such models-Grounding DINO and SAM 2-to construct a pipeline for zero-shot skin segmentation in dermatology. We evaluated our approach on two clinical skin photography datasets comprising 27,378 images. Based on a manual rating protocol, 77.1% of the segmentations were deemed acceptable, demonstrating robustness in handling real-world clinical photographs. Our results highlight the potential of open-world foundation models to address a challenging problem in dermatology with minimal human involvement.

Accurate segmentation of seizure phases in intracranial EEG is essential for characterizing seizure dynamics and supporting presurgical evaluation in drug-resistant focal epilepsy. This study examines whether a semi-supervised changepoint detection framework can reliably delineate ictal onset, intra-ictal transition, and seizure termination. A three-phase segmentation pipeline integrates multivariate envelope-based features, including root mean square amplitude, relative bandpower in the theta (4-8 Hz), alpha (8-13 Hz), beta (13-30 Hz), and gamma (30-80 Hz) bands, line length, and spectral entropy, with the Pruned Exact Linear Time algorithm. Features were extracted from sliding windows whose lengths and phase-specific weights were optimized using nested leave-one-subject-out cross-validation with Optuna. To ensure length invariance, analysis windows were randomly extended by 5-30 s before seizure onset and after seizure termination using real pre- and post-ictal data. Performance was evaluated on 179 seizure-onset-zone bipolar channels across 32 seizures from 10 patients. Mean absolute errors were s for seizure onset, s for intra-ictal transition, and s for seizure termination. Detection accuracies within s were 71.6% for onset, 60.0% for transition, and 75.0% for termination. Phase-specific feature importance analysis revealed distinct and evolving contributions of amplitude-, spectral-, and complexity-based measures across seizure phases. The proposed framework achieves temporal precision comparable to reported inter-rater reliability (Cohen's -0.69) and provides an interpretable, data-driven approach for comprehensive seizure phase characterization, with potential utility in clinical decision-making.

DCDGNet: Dual-frequency cross-feature diffusion GAN for single fusion sonar image generation in exposed subsea pipeline inspection.

AI-driven road inspection with SUD-ROAD: High-resolution LiDAR benchmark and a novel cross-dimensional semantic segmentation pipeline

Large Language Models (LLMs) are increasingly applied to automate complex tasks, but their potential for generating automated medical image segmentation pipelines remains largely underexplored. We present a systematic evaluation of open- and closed-source LLMs in generating U-Net–based segmentation frameworks across six diverse 2D medical image datasets, spanning endoscopy, fluoroscopy, dermoscopic photography, MRI, and fundus imaging. Building on an earlier study, we analyze state-of-the-art 2025 reasoning-enabled models and compare them to non-reasoning LLMs and a strong nnU-Net v2 baseline. Compared to their 2024 predecessors, the 2025 models demonstrated marked improvements in robustness, code quality, and segmentation accuracy across modalities. Our results show that reasoning-augmented LLMs achieve faster convergence, fewer execution errors, and higher Dice scores, while complex datasets with fine structures (e.g., retinal vessels) and volumetric data remain challenging. We also confirmed robustness under repeated runs by comparing one reasoning and one non-reasoning model from the same family, where despite GPT-4o’s consistent, template-like code outputs under multiple runs as the non-reasoning model, GPT-o4-mini-high showed significantly lower run-to-run variability in validation loss and tighter Dice score distributions, demonstrating that chain-of-thought reasoning markedly improves both accuracy and stability. These findings highlight the potential of reasoning-enabled LLMs to automate segmentation workflows with high accuracy and explainability, paving the way for their integration into medical imaging pipelines. Our code is available at <a href='https://github.com/ankilab/LLM_based_Segmentation.git'>https://github.com/ankilab/LLM_based_Segmentation.git<a>

Glioblastoma (GBM) lethality arises from aggressive invasion and diffuse infiltration of brain tissue. Conventional GBM preclinical models often fail to predict clinical therapeutic efficacy because they do not recapitulate the pathological extracellular matrix (ECM) cues that drive tumor invasion. Here, we present an ECM mimetic 3D platform using a fibrin scaffold to recapitulate the hemorrhagic, pro-thrombotic tumor microenvironment characteristic of high-grade gliomas. This fibrin scaffold induces a pro-invasive phenotype in GBM spheroids by upregulating proliferation/cell cycle- ( MYC, FOXOM1, CCND1 ) and invasion-associated-( CTSS, FOXM1, CCND1 ) genes. Traditional cell morphology quantification methods (e.g., circularity) distil complex shapes into singular metrics and cannot capture the nuances of invasion. To address this limitation, we have applied a deep-learning segmentation pipeline (MARS-Net) and high-content morphodynamic descriptors. By using the Preserving Heterogeneity (PHet) algorithm, the 3D platform accurately classifies invasiveness levels and captures the invasion-inhibitory effects of potential repurposable drug candidates. We demonstrate that our model can predict a spheroid's long-term invasive fate with high accuracy using only partial image sets from early time-points, rather than the complete time-course images. Our work presents an in vivo -like, scalable 3D platform integrated with a quantitative high-throughput pipeline to elucidate GBM invasion mechanisms and to evaluate anti-invasive compounds.

PET imaging with [18F]F-DOPA shows great promise for assessing paediatric gliomas. Manual tumour delineation and parameter extraction are time-consuming and prone to inter-operator variability. We evaluated whether a deep learning model, leveraging transfer learning from adult glioma datasets, could enable a fully automated pipeline for tumour segmentation and PET parameter extraction. Static and dynamic parameters were compared across three approaches: (i) automatic vs semi-automatic, (ii) automatic vs manual, and (iii) manual vs. semi-automatic. Data from 103 paediatric patients (median age 11years; 54 females, 49 males) with static and/or dynamic [18F]F-DOPA PET scans (2011-2024) were retrospectively included for fine-tuning the deep learning model. Statistical and survival analyses were performed on 90 subjects; dynamic analysis included 32 patients. The best model achieved a Dice score of 0.82 ± 0.11 and was integrated into the pipeline for extracting static and dynamic indices. Automatic Tumour-to-Striatum ratio showed high reproducibility across comparisons ((i) p = 0.660, (ii) p = 0.342, (iii) p = 0.639), while Tumour-to-Background differed significantly when comparing manual delineations (p < 0.01). Dynamic parameters demonstrated good reproducibility with the automatic method (p > 0.05). Importantly, both automated static indices correlate significantly with tumour grade, with the overall and progression-free survival (p < 0.05). Transfer learning enabled a fully automatic [18F]F-DOPA PET pipeline for paediatric gliomas, providing reproducible static and dynamic parameter extraction and correlating with clinically relevant outcomes. This approach reduces operator dependence and streamlines analysis, supporting potential integration into routine clinical practice.

Background: Chest X-ray (CXR) image classification remains a critical tool in COVID-19 and pneumonia diagnosis. While segmentation of lung fields is commonly assumed to improve deep learning classification performance, recent evidence suggests that segmentation may also remove clinically relevant contextual cues. Attention mechanisms, particularly those that combine spatial and channel information, have shown potential to enhance model focus and generalizability. Objective: This study investigates the effectiveness of the proposed Self-Adaptive Convolutional Block Attention Module (SA-CBAM) in improving CXR image classification performance when tested on both segmented and unsegmented images. In addition, the benefit of hybridization of Convolutional Neural Network (CNN) with a Long Short-Term Memory (LSTM) network is evaluated. Methods: A U-Net segmentation model was used to extract lung regions from the COVID-QU-Ex dataset. Multiple model variants were implemented and compared, including a baseline CNN, CNN models integrated with several channel–spatial attention mechanisms, and the proposed SA-CBAM. In addition, hybrid architectures combining CNN with LSTM networks were evaluated. All models were assessed on both segmented and unsegmented CXR pipelines using accuracy, recall, specificity, F1-score, and Matthews Correlation Coefficient (MCC) as primary performance metrics. Results: Models trained on unsegmented CXR consistently outperformed those trained on segmented images. The proposed CNN-SA-CBAM improved baseline CNN performance from 88.42% to 90.08% accuracy and from 82.82% to 85.36% MCC when trained on unsegmented data. Further hybridization with an LSTM network produced the highest performance, where the CNN-SA-CBAM-LSTM achieved 99.90% accuracy and 99.85% MCC on unsegmented CXR. Despite segmentation producing visually clearer attention maps, classification performance was lower, suggesting potential loss of subtle contextual cues outside the segmented lung boundaries. Conclusion: The findings demonstrate that SA-CBAM significantly enhances CXR classification performance, particularly when applied to unsegmented images and further combined with LSTM. This study challenges the common assumption that segmentation always improves classification accuracy and highlights the importance of preserving contextual information. Future work will focus on adaptive or soft segmentation strategies that retain peripheral cues while preserving anatomical interpretability.

Carotid artery segmentation is critical for determining the degree of vascular disease, and for recommending treatment options. Early detection of carotid atherosclerosis is critical for preventing stroke. Stroke-related brain damage can cause deficits in speech or vision, and large strokes can be fatal. However, automatic segmentation of the carotid artery lumen remains difficult due to the low quality of US images, and the existence of stenosis, jugular veins, and abnormalities in carotid images. This article presents a hybrid pipeline for segmenting both carotid transverse and longitudinal lumens without any user interaction. This hybrid pipeline starts with automatically localizing the carotid artery lumen in the transverse and longitudinal sections via YOLOv11n. Then, a multistage preprocessing framework was applied to the transverse section before its lumen was segmented by the active contour. For the longitudinal section, an automated padded mask was generated to guarantee reliable initialization for Chan-Vese level-set evolution. A paired t test validated the relevance of the proposed modules (p < 0.0001). The proposed multiphase segmentation pipeline achieved a Dice index and accuracy of 94.9% and 97.7%, respectively, for the longitudinal section and 90.8% and 99.6%, respectively, for the transverse section. A comprehensive ablation analysis has shown that numerical stability depends on the YOLOv11n localization phase. The system attained near real-time inference for the carotid transverse section (< 1s) despite being evaluated on low-end hardware, demonstrating its computational efficiency and promise for clinical integration.

Bone tumor detection from X-ray images is challenging due to noise, low contrast, and irregular tumor boundaries that complicate precise segmentation. This study proposes a lightweight Relational YOLO–SegNet framework integrating an Optimized Savitzky–Golay Digital Filter (OSGDF), an ROI-restricted Relational Transformer Block (RTrB), and Fire Hawk Election Optimizer (FHEO)-based hyperparameter tuning. The proposed framework operates as an ROI-guided detection–segmentation pipeline, where YOLOv8 first localizes tumor regions, after which a Relational YOLO–SegNet model performs precise pixel-level segmentation. Image-level classification of normal versus tumor cases is subsequently derived from the segmented regions, making segmentation the primary objective of the framework. The unique contribution of OSGDF lies in its adaptive parameter selection using Tunicate Swarm Optimization, which improves noise suppression while preserving edge sharpness; this resulted in an SNR improvement from 21.4 dB to 29.6 dB, enhancing boundary delineation prior to segmentation. The proposed model applies relational attention only to YOLO-detected tumor regions rather than the entire image token space, reducing computational complexity while maintaining long-range contextual modeling. The framework contains 12.3 million trainable parameters, fewer than conventional encoder–decoder architectures such as UNet and Mask R-CNN. Experiments conducted on a publicly available dataset of 809 X-ray images (421 normal, 388 tumor) with expert-provided pixel-level annotations achieved 98.5% accuracy, 98.32% precision, 98.83% sensitivity (recall), 98.21% specificity, 98.57% F1-score, 97% Dice score, 97.1% Jaccard index, and an AUC of 0.981 under five-fold cross-validation (98.5 ± 0.3%). Statistical analysis confirmed that improvements over baseline models were significant (p < 0.05). The model achieved an inference time of 48 ms per image on an NVIDIA RTX 3090 GPU (24 GB VRAM), demonstrating computational efficiency suitable for resource-constrained deployment scenarios. While the results indicate strong dataset-specific performance, external multi-institutional validation is required before clinical translation. The proposed framework may serve as a potential research-support tool for automated bone tumor analysis using X-ray imaging.

Frailty in oncology is a major determinant of treatment toxicity and survival, and is often framed primarily as a muscle problem. Adipose tissue, however, is an active endocrine and metabolic organ, and its glycolytic activity on positron emission tomography/computed tomography with fluorodeoxyglucose (18 F-FDG PET/CT) may capture physiological vulnerability that is not reflected by body composition alone. We investigated the association between adipose glycolytic activity and frailty in older adults with solid tumours. We prospectively enrolled 104 adults (≥ 50 years) with solid malignancies (median age 63.5 years) who underwent clinical whole-body 18 F-FDG PET/CT and a comprehensive geriatric assessment on the same day. At the L3 level, adipose area and metabolic activity (SUVmean and rSUVmax; SUVp95-derived and reference-normalized hereafter referred as SUVmax for simplicity) were quantified using a deep learning segmentation pipeline (TotalSegmentator) with strict exclusion of visceral structures to isolate adipose signal. Frailty was assessed using the Clinical Frailty Scale (CFS) and the FRAIL Scale. Total adipose area did not differ between frail and non-frail phenotypes (425.64 vs. 424.38 cm², p = 0.45). In contrast, adipose glycolytic activity was significantly higher in frail patients (SUVmean 0.30 vs. 0.20, p < 0.001). In multivariable logistic regression adjusted for age and sex, each 0.1-unit increase in adipose SUVmean was associated with 1.78-fold higher odds of frailty (95% CI 1.10–2.88, p = 0.002). Associations were directionally consistent across both frailty instruments. Adipose hypermetabolism on 18 F-FDG PET/CT, despite low absolute SUV values, appears to track frailty independently of adipose quantity, supporting a “fat heat-up” phenotype as a marker of diminished physiological reserve. Routine oncologic PET/CT may therefore provide an opportunistic, imaging-derived frailty signal that can precede overt morphological deterioration in body composition. Frailty is a high-impact and frequently under-recognised driver of treatment intolerance and adverse outcomes in oncology, yet comprehensive geriatric assessment remains resource-intensive and inconsistently implemented. In this prospective cohort, adipose tissue was quantified using a fully automated Total Segmentator-based pipeline, providing an objective and reproducible measure of adipose FDG uptake at the L3 level. This single quantitative feature, already embedded in routine 18F-FDG PET/CT, aligned with frailty across two independent frailty scales and multiple geriatric assessment domains, independent of age and sex. If externally validated, adipose FDG uptake could function as an opportunistic imaging-derived frailty flag to trigger earlier geriatric input, support treatment individualisation, and enable physiologic risk stratification in trials and real-world practice.

Background/Objectives: Translating brain volumetric biomarkers to individual-level Alzheimer's disease (AD) diagnosis remains challenging due to difficulty interpreting raw volumes without longitudinal monitoring or matched controls. We tested a classification model using population-referenced volumetric percentiles to distinguish AD from cognitively normal (CN) subjects and evaluated its generalization across independent cohorts. Methods: Brain volumes from 95 regions were extracted using an automated segmentation pipeline and converted to age and sex adjusted percentiles using a reference population (N = 1833). A logistic regression classifier was trained on ADNI subjects (N = 873; AD = 183, CN = 690) split into training (60%), validation (20%), and test (20%) sets. The model was evaluated on two independent validation datasets: the held-out ADNI validation set and an external Korean cohort (N = 72; AD = 36, CN = 36) acquired with different scanner protocols and demographic characteristics. Results: The model achieved excellent discrimination across all evaluation sets: ADNI validation (AUC = 0.963, accuracy = 90.3%), ADNI test (AUC = 0.960, accuracy = 89.7%), and Korean external validation (AUC = 0.981, accuracy = 87.5%). The minimal validation gap (0.018) demonstrated robust generalization. Positive coefficients for ventricular regions reflected AD-associated atrophy patterns, while negative coefficients for medial temporal structures indicated their contribution within multivariate patterns distinguishing AD from normal aging. Conclusions: Population-referenced brain volumetric percentiles enable accurate AD classification with robust generalization across populations and scanner protocols. By contextualizing individual brain structure relative to normative populations while accounting for age and sex, this approach demonstrates potential for clinical translation as an accessible neuroimaging-based diagnostic tool.