共找到 20 条结果
•Automated scoring on HSAT devices underestimate REI compared to manual scoring.•Automated scoring underestimated OSA severity in up to 44.5% of moderate cases.•Misclassification occurs near clinically relevant severity thresholds.•Hypopnea index was significantly lower with automated compared to manual scoring.•Findings highlight targets for improving automated and AI scoring.
The Evans index (EI) is widely used to assess ventricular enlargement and support external ventricular drainage (EVD) decision-making after subarachnoid hemorrhage (SAH), but manual measurement is time-consuming and subject to inter-reader variability. The reliability of automated EI measurement in acute SAH remains insufficiently validated. In this retrospective cohort study, admission non-contrast CT scans from 364 patients with spontaneous SAH were analyzed using TotalSegmentator (TS), an open-source nnU-Net-based deep learning pipeline, to generate automated EI measurements. Each scan underwent two independent inference runs, and two neurosurgical experts independently performed manual measurements. Agreement between TS and expert measurements was evaluated for continuous EI values and EI > 0.30 classification. External ventricular drainage (EVD) placement was used as a pragmatic endpoint reflecting contemporaneous clinical decision-making. Prespecified subgroup analyses excluded frontal horn hematoma or frontal horn periventricular edema. Multivariable logistic regression assessed the association between EI and EVD placement after adjustment for key clinical covariates. TS demonstrated excellent reproducibility between repeated inference runs (ICC = 0.996, 95% CI 0.996-0.997). Agreement between expert readers was high (ICC = 0.983, 95% CI 0.978-0.988). Between-method agreement between TS and expert EI measurements was good in the overall cohort (ICC = 0.76, 95% CI 0.73-0.81) and improved after exclusion of frontal horn hematoma (ICC = 0.87, 95% CI 0.85-0.89). For EI > 0.30 classification, TS identified more positive cases than expert assessment (29% vs. 17%), with moderate Cohen's kappa (0.57, 95% CI 0.54-0.60). TS-derived EI demonstrated discrimination for EVD placement (AUC = 0.75, 95% CI 0.73-0.79), approaching expert-derived EI (AUC = 0.80, 95% CI 0.78-0.83). After covariate adjustment, TS-derived EI remained independently associated with EVD placement (adjusted OR = 1.09, 95% CI 1.03-1.17; p = 0.009). Automated EI measurement using TS provides reproducible and clinically informative assessment of ventricular enlargement on CT in acute SAH. Although threshold-sensitive disagreement occurred near EI = 0.30, automated EI showed meaningful agreement with expert assessment and remained independently associated with contemporaneous EVD decision-making. Further SAH-specific refinement may improve robustness in hemorrhage-related ventricular distortion.
The diagnosis of mycosis fungoides is difficult and often delayed, exacerbated by the limitation of conventional immunohistochemistry to analyze only 1 antigen per tissue section, often necessitating repeat biopsies and extensive workups. We sought to validate a high-throughput multiplex immunofluorescence method, coupled with computer-automated image analysis, to generate comprehensive immunophenotyping data from a single formalin-fixed, paraffin-embedded biopsy. We applied an 11-biomarker multiplex immunofluorescence panel across 18 archived skin specimens (9 mycosis fungoides/TCR clonality positive and 9 control/TCR clonality negative). Initial validation confirmed that multiplex immunofluorescence antigen expression and spatial localization were concordant with those of sequential immunohistochemistry-stained sections. Whole-slide image stacks were analyzed using both computer-assisted and computer-automated pipelines. Both methods successfully delineated immunophenotypic differences. Mycosis fungoides specimens showed a significant expansion of hematopoietic cells, atypical T-lymphocytes, and proliferative T-lymphocytes when compared with controls. Our results validate the potential of multiplex immunofluorescence to obtain comprehensive, high-dimensional diagnostic information from a single tissue section. Integration with computer-automated analysis offers a scalable, high-throughput platform that can significantly aid in the timely and accurate diagnosis of cutaneous lymphomas.
Manual kinematic analysis in swimming is labor-intensive and often lacks the immediacy required for elite training. This study presents an automated vision-language framework to deliver real-time kinematic profiling and coaching diagnostics. A robust object detection pipeline using YOLOv11 was developed, incorporating a splash-injection strategy to handle aquatic occlusion. Kinematic metrics (velocity, distance) were extracted via homography transformation. To automate pedagogical feedback, the DeepSeek-V3 Large Language Model was integrated to interpret these metrics and generate structured coaching reports. The proposed method achieved a mean Average Precision (mAP@0.5) of 94.64% in dynamic water conditions. The system successfully tracked swimmers despite turbulence and accurately synthesized quantitative data into natural language assessments of pacing and fatigue. This open-source framework significantly reduces the manual burden of performance analysis. By combining computer vision with automated reporting, it offers a scalable, objective tool for daily swim training and technical evaluation.
Whale sharks (Rhincodon typus) are migratory species known to form temporary coastal aggregations in regions of high plankton density, such as Bahía de los Ángeles (BLA), Mexico. Due to various anthropogenic threats, the species is classified as vulnerable, underscoring the need for effective and non-invasive monitoring strategies. This study explores the integration of drone-based remote sensing with deep learning (DL) techniques for the automated recognition and tracking of whale sharks. In October 2023, 59 drone flights-42 from boats and 17 from land-were conducted, resulting in 23 recorded sightings. Two DL-based approaches were implemented: DeepLabCut (DLC), based on pose estimation, and multi-scale patch (MSP), based on image region classification. Both strategies employed convolutional neural networks, with data augmentation and transfer learning techniques to enhance performance. The most effective model, MSP with multi-size and overlapping patches, achieved a macro F1 score of 0.91 and a covered area metric of 0.65. Statistical analysis indicated that environmental and operational factors-including turbidity, solar glare and external agents-had minimal impact on predictive performance. These results highlight the potential of DL-enhanced drone monitoring for scalable, accurate and non-invasive recognition and tracking of whale sharks, supporting ecological research and conservation efforts in BLA and comparable marine environments.
Arbuscular mycorrhiza (AM), a symbiosis between land plants and fungi that enhances plant mineral nutrition, has potential for application in sustainable agriculture. The extent to which plants benefit from the symbiosis depends on the plant-fungal genotype combination, providing opportunities for breeding symbiosis-optimized crops. Here, we analyzed several plant growth traits and shoot mineral element accumulation of 15 genetically diverse maize inbred lines in response to AM. Plants were grown in an automated phenotyping system, in which randomization produced reproducible shoot surface growth curves. We observed strong variation in AM responsiveness of shoot and root growth in a phosphate-poor alkaline substrate. This substrate caused stunted growth in non-inoculated control plants and a 165% to 633% increase in plant biomass in AM plants. We observed variation among the maize lines in the AM-mediated increase in total shoot content of 24 mineral elements plus phosphate and sulphate. While nitrate accumulated in control plants, it was not detectable in the shoots of AM plants. Growth traits and accumulation of mineral elements were not uniformly correlated across the 15 maize lines. An exploratory ridge-regression analysis showed that RAM was only selectively predictable across traits, indicating that RAM is mostly trait-specific rather than uniformly coordinated across traits.
With the widespread implementation of automated hematology systems, Reflex testing plays a critical role in result verification. However, overly sensitive Reflex rules may lead to excessive repeat testing, increasing workload and prolonging sample turnaround time. This study optimized Reflex rules to improve workflow efficiency while maintaining analytical safety and accuracy. A retrospective analysis was performed on 78,364 CBC results generated by the Sysmex XN-9100 hematology automation line in June 2025, including 43,758 outpatient and 34,606 emergency/ward samples. Reflex rules involving LW, PLT-F, WPC, and RET channels were optimized by differentiating initial and follow-up visit samples, introducing a 7-day historical result fluctuation safety range, and simplifying IP message combinations. Middleware simulation compared total Reflex rates and special-channel repeat counts before and after optimization. Existing autoverification logic, critical alert rules, manual review workflow, and high-risk disease-related repeat testing rules were retained. After optimization, total Reflex rates decreased significantly in both work areas (p < 0.01), from 16.03% to 14.36% in the outpatient area (relative reduction, 10.43%) and from 24.38% to 15.00% in the emergency/ward area (relative reduction, 38.47%). In the emergency/ward area, PLT-F, WPC, and RET repeat counts decreased by 44.86%, 49.19%, and 61.89%, respectively. XN automation line Reflex rule optimization reduced unnecessary repeat testing through initial/follow-up visit differentiation and historical fluctuation thresholds while retaining core risk-interception rules, supporting improved hematology workflow efficiency and reduced reagent consumption.
To analyze the spatial and epidemiological characteristics of out-of-hospital cardiac arrest (OHCA) in Deyang, China (2021-2023), quantify the supply-demand coverage gap in current automated external defibrillator (AED) placement, and evaluate a geospatial optimization strategy embedding AEDs in 24/7 accessible outdoor cabinets at primary healthcare institutions (PHCIs). We integrated data on OHCA cases, existing AEDs, PHCIs, and demographics. Temporal variations were assessed via descriptive epidemiology, while spatial heterogeneity was characterized via Global/Local Moran's I and Getis-Ord Gi* hotspot analysis. A network analysis model integrating the Gaussian Two-Step Floating Catchment Area (Ga2SFCA) method and the Maximal Covering Location Problem (MCLP) was constructed using PHCIs as candidate nodes based on road network distance. Optimization was simulated under two scenarios: complete redistribution and incremental expansion. The study included 3273 OHCA cases. Significant temporal variations were observed, with peak incidence during morning hours and public holidays. Spatial autocorrelation indicated intensifying OHCA clustering from 2021 to 2023 (Moran's I: 0.38-0.49, P < 0.01). Furthermore, spatial analysis revealed a stable 'core-periphery' structure, with Jingyang District identified as a consistent high disease burden hotspot. A marked supply-demand coverage gap was observed: residential areas accounted for 82.22% of cases but only 6.29% of AED deployments. Existing AEDs covered only 31.87% of historical OHCAs within a 500-m road network distance. The new Optimized Model for Network Analysis demonstrated that achieving >50% coverage required 100 additional AEDs under a complete redistribution strategy, versus 200 under an incremental expansion strategy. The distribution of OHCA incidence exhibits distinct epidemiological patterns, significant spatial clustering, and a marked disconnect from current AED placement. A geospatial optimization strategy leveraging the extensive network of PHCIs significantly enhances coverage efficiency. This "PHCI-embedded" model provides a cost-effective framework for establishing resilient cardiac arrest response systems in resource-constrained urban settings.
Infectious lung diseases pose a serious public health risk; can cause long-term conditions like fibrosis, loss of lung function and chronic lung diseases like COVID-19, pneumonia and tuberculosis. Persistent lung abnormalities can result from these complications and need to be accurately detected and analysed. The deep ensemble learning technique was utilized to extract high-quality images from raw computed tomography (CT) scan images by incorporating noise detection and reduction techniques. In this study, BM3D denoising technique has been used for noise reduction in the post-infectious CT scan image to improve its quality by using the automated detection of noise. The main objective of this research is to improve the image quality in the pre-processing stage using various noise detection methods such as Blur Laplacian, Noise Entropy, Noise Power Spectrum Density and Wavelet Transform. Each noise detection technique produces the amount of noise detected in terms of numeric variance. Based on the highest numeric variance, the noise type is detected automatically. Then, the BM3D denoising technique was applied to yield an effective image for further processing. The proposed system validates noise reduction in terms of PSNR, SSIM, and edge preservation index. Thus, the results provide a significant enhancement in image clarity and structural clarity to support downstream diagnostic tasks, supporting its potential utility as a pre-processing step for clinical and research applications.
Performance of deep learning-based models for tumour tracking in magnetic resonance-guided radiotherapy hinges on high-quality labelled data. Medical images are prone to annotation errors, making data cleaning essential. We propose an automatic data cleaning tool based on the foundation model Segment Anything 2, which incorporates temporal information from cinematic magnetic resonance imaging to detect annotation errors and generate corrected contours, minimising manual effort involved in data cleaning. Expert validation by two radiation oncologists showed a preference for corrected contours over the original manual contours. Corrected contours (DSC 0.95 ± 0.01) surpassed interobserver variability (0.88 ± 0.02) on a dataset annotated by five observers.
The primary objective of this study was to develop and validate SCC-Net, a lightweight GhostNet-CBAM model, for the early and accurate detection of cutaneous squamous cell carcinoma (cSCC) from dermoscopy images, addressing the limitations of resource-constrained clinical settings. We developed SCC-Net, a lightweight GhostNet-CBAM model trained on 5150 public dermoscopy images (628 cSCC) formulated as a binary classification task (cSCC vs. non-cSCC). Using five-fold cross-validation and external testing against five CNN baselines, evaluating accuracy, recall, F1, AUC, and Grad-CAM interpretability. SCC-Net achieved 96.6% accuracy, 88.7% recall, and 0.986 AUC internally, and 94.3% accuracy, 93.8% recall, and 0.986 AUC externally, outperforming all comparables with only 1.59 M parameters. Grad-CAM heatmaps aligned with expert lesion focus. SCC-Net delivers accurate, explainable, and low-cost cSCC detection, promising for teledermatology and resource-limited settings.
Control efforts against schistosomiasis are hampered by the subjective interpretation of the point-of-care circulating cathodic antigen (POC-CCA) urine test, which limits diagnostic consistency. We developed MEDSCAN (Mobile-Enabled Diagnostics for Schistosomiasis Control Analytics), a mobile application that uses smartphone imaging and computer vision to automate POC-CCA interpretation. In a multi-site laboratory evaluation across the USA, the Netherlands, and Kenya, we compared MEDSCAN to visual G-Score interpretation and a benchtop lateral flow reader (LFR). All three methods produced clear concentration-response relationships, with normalized machine-based metrics achieving AUROC ≥ 0.90. MEDSCAN demonstrated excellent inter-user reproducibility (intra-class correlation coefficients exceeding 0.94) and substantial agreement with both visual and LFR interpretations across sites. Some device-to-device variability was observed, attributable to differences in smartphone camera hardware and image processing; however, binary diagnostic outcomes remained robust across a heterogeneous set of smartphones. These results establish operational positivity thresholds for MEDSCAN-based on test-line signal alone or normalized metrics-suitable for direct implementation in field surveillance workflows. A large-scale field study is underway to evaluate MEDSCAN under routine POC-CCA surveillance conditions in endemic settings.
To develop a reproducible generative artificial intelligence (GenAI)-driven workflow for periodontitis risk stratification using systemic and demographic indicators, and to validate its ability to identify well-established predictors in resource-limited settings. This retrospective study analyzed data from 416 dental hospital patients. Using systematic prompt engineering, GenAIwas employed to automate data preprocessing, correlation analysis, and development of six machine learning models (namely, Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, Support Vector Machine, and K-Nearest Neighbors) to predict periodontitis severity. Severe periodontitis was defined as Community Periodontal Index (CPI) score of 4. Model validation was performed using an 80-20 data split and fivefold cross-validation and McNemar's Test. The GenAI-driven pipeline successfully automated the data analysis workflow. Models achieved modest discriminatory power using systemic indicators alone (AUC 0.48-0.57). Logistic Regression demonstrated the most balanced performance (72% accuracy, 74% F1-score), while Support Vector Machine (SVM) showed superior sensitivity (89%) for screening severe cases. Feature importance analysis identified age (score = 0.233) and blood sugar level (score = 0.209) as the strongest predictors, consistent with established periodontal risk factors. Notably, composite systemic risk scores exhibited a stronger correlation with periodontitis severity than any individual health parameter. While systemic indicators alone provided limited diagnostic precision, the GenAI-driven workflow effectively automated data process with end-to-end model development. The high sensitivity of the SVM model suggests potential utility as a preliminary screening tool to flag at-risk individuals for prioritized clinical examination, particularly in settings where dental radiography is unavailable. This research demonstrates the potential of GenAI to facilitate efficient and interpretable risk stratification rather than definitive diagnosis. The workflow provides a replicable, privacy-preserving framework that lowers the technical barrier to applied machine learning in resource-limited periodontal care.
Arthritis is a widespread musculoskeletal disease, which is characterized by inflammation, pain, stiffness, and progressive loss of joint mobility and has a significant impact on quality of life across the globe. To achieve early clinical intervention and better patient outcomes, it is important to detect arthritic changes early. Nonetheless, traditional radiographic diagnosis is largely dependent on the interpretation of experts, which can lead to variability and timeliness. The research seeks to create an automated mechanism of precise and timely identification of arthritis through the use of deep learning. The framework based on the Convolutional Neural Network (CNN) was suggested to support automated detection of arthritis by using knee X-ray images. The model has been trained and assessed on a publicly available Kaggle dataset of 4414 radiographic images of normal and arthritic knee conditions. Image normalization and resizing were performed in advance to ensure image consistency. The CNN architecture was composed of several convolutional layers to obtain features, and fully connected layers to perform binary classification. Accuracy, precision, recall, confusion matrix, and Receiver Operating Characteristic (ROC) were the performance measures used to evaluate the model's performance. Grad-CAM visualization was added to make the interpretation more interpretable. The overall classification accuracy of the proposed model was 95%, and the precision and recall were 0.98 and 0.93, respectively. The confusion matrix and ROC analysis indicated a high level of discriminative ability in differentiating arthritic and normal knee conditions. Grad- CAM visualizations were used successfully to highlight relevant areas in radiographs that have been used to predict the model. The results show that deep learning-based analysis of knee radiographs can offer reliable and efficient detection of arthritis. The interpretability methods, like Grad-CAM, improve trust and transparency in model predictions that can be more applicable to clinical practice. This methodology overcomes the weaknesses of subjective interpretation of conventional diagnostic approaches. This paper has shown that a CNN-based system could be used to identify arthritis based on knee X-rays with high accuracy and reliability. The proposed system can be used to support clinicians in musculoskeletal imaging workflows and can be integrated into computeraided diagnostic systems to assist in the early screening and better patient care.
Digital pathology continues to transform the daily routine of pathology, in terms of the increasingly automated laboratory and in the diagnostic paradigm through the adoption of artificial intelligence (AI) tools to support diagnosis-computational pathology. The reliability and performance of these tools depend on the whole-slide image (WSI) quality being guaranteed a priori. Pre-analytical quality control step that underpins this guarantee, and artifact detection remains largely qualitative and is frequently overlooked in routine digital pathology. This operational feasibility study evaluated whether an adaptation of GrandQC, an open-source AI tool, enables automated, quantitative artifact assessment of a complete single-day biopsy workload from a high-throughput digital pathology laboratory, analyzed retrospectively. A random biopsies day of 2025 at Centro de Anatomia Patológica Germano de Sousa (CAPGS) was selected as a sample to test the performance of GrandQC on the WSI generated (n = 544) in order to simulate the daily workflow. A script was created to quantify the pixels corresponding to the type of artifact automatically, creating an Excel file for registering and statistical analysis. Analysis took a median of 24 s per WSI, detecting a median of 1.46% of tissue area with some type of artifact. Dark Spots and blurring areas were the most representative detected artifacts. GrandQC is a valuable tool in the quantitative quality control of biopsies tissue, allowing quick evaluation, signaling types of artifacts, and identifying cases that need to be reviewed before being handed over to the pathologist allowing the recognition of opportunities to improve laboratory histology quality and precision medicine.
Biofouling-communities of organisms that grow on submerged hard surfaces-creates pathways for the spread of invasive marine species and diseases. To manage this risk, international vessels are increasingly required to demonstrate effective biofouling management, which in turn necessitates underwater inspections and the analysis of large volumes of hull imagery to verify biofouling status. Automated assessment with computer vision can streamline this process. This work shows how the interpretable Component Features (ComFe) approach, combined with a DINOv2 Vision Transformer (ViT) foundation model, can address this challenge efficiently and effectively. ComFe achieves competitive performance in comparison to previous non-interpretable CNN methods, with fewer parameters and greater transparency-highlighting which image regions and training examples drive classifications. All code, data, and model weights are publicly released.
The increasing clinical demand for electroconvulsive therapy (ECT) highlights specific challenges in anaesthetic management, particularly the reliance on continuous mask ventilation. Bilevel positive airway pressure (BPAP) provides automated, hands-free ventilatory support and may enhance efficiency. This study aims to compare the impact of BPAP versus face mask ventilation on procedural turnover time during ECT. In this single-blind, randomized controlled trial, 110 patients were enrolled with a history of at least two prior ECT sessions. Participants were randomly allocated to receive either BPAP or face mask ventilation. The primary outcome was the turnover time, defined as the duration from anaesthesia induction to entry into the post-anaesthesia care unit (PACU). Secondary outcomes included the total ECT duration, breathing and consciousness recovery times, incidence of hypoxemia (SpO2 < 94% for > 5 s), adverse events and operator-centered measures, including upper limb discomfort, anxiety and satisfaction scores. The BPAP group demonstrated a significantly shorter median turnover time ccompared to the face mask group (332 [286-415] s vs. 426 [348-472] s; median difference 84 s, 95% CI 44-119, P < 0.001). No significant differences in the total ECT duration, breathing recovery time, consciousness recovery time, or incidence of hypoxemia. The incidence of anaesthesiologist-reported upper limb discomfort was lower in the BPAP group (P < 0.001), and operator satisfaction scores were higher (P < 0.001). In patients with prior ECT experience, BPAP safely reduced procedural turnover time and significantly decreased the physical burden on anaesthesiologists compared to face mask ventilation. BPAP represents a viable strategy to improve workflow efficiency and operator comfort in high-volume ECT settings. ChiCTR2100051837, the Chinese Clinical Trial, 20-211-006. Research registration number: ChiCTR2100051837, the Chinese Clinical Trial Registry. IRB number: [2021] 02-203-01, the Research Ethics Committee of the Third Affiliated Hospital of Sun Yat-sen University, Guangzhou, China.
The relevance of the study is as follows. Traditional methods of pedagogy do not allow objective and continuous monitoring of the emotional state of each student. The teacher usually relies on subjective observations - facial expressions behavior in a group - which may be inaccurate. At the same time modern technologies provide new opportunities: cameras and sensors backed by AI algorithms are able to automatically recognize facial expressions voice intonations and other signs of an emotional state. There are already commercial solutions for facial emotion recognition which have been used in marketing and security. Their potential application in education is an important topic as automated emotion analysis could provide educators with objective data on how engaged a group is whether students are overworked and who is experiencing difficulties or stress. It is the problem of mass learning fatigue and emotional burnout according to empirical studies that makes the introduction of such systems modern and practically significant and is the topic of this article. The article notes that the introduction of AI systems that can monitor both behavioral and physiological indicators opens up new opportunities for objective registration of cognitive states. For example sensors and wearable devices can measure heart rate heart rate variability brain activity and other parameters related to the level of concentration and mental stress. Computer vision can track eye gaze direction blinking frequency and posture which indirectly indicates the level of attention or fatigue. The study presents the results of a psychometric testing study using validated tools for emotional involvement in the educational process and a study of the factors of burnout and knowledge fatigue among full-time students at the Novorossiysk branch of the Financial University under the Government of the Russian Federation. The sample consisted of 125 respondents. Традиционные методы педагогики не позволяют объективно и непрерывно отслеживать эмоциональное состояние каждого студента. Преподаватель обычно опирается на субъективные наблюдения — выражения лиц поведение в группе — которые могут быть неточными. Современные технологии дают новые возможности: камеры и датчики подкреплённые алгоритмами ИИ способны автоматически распознавать мимику голосовые интонации и другие признаки эмоционального состояния. Уже существуют коммерческие решения для распознавания эмоций по лицу которые применяются в маркетинге и безопасности. Проблема массовой учебной усталости и эмоционального выгорания делает внедрение подобных систем современным и практически значимым и является темой данной статьи. В статье отмечено что внедрение ИИ-систем способных проводить мониторинг как поведенческих так и физиологических индикаторов открывает новые возможности для объективной регистрации когнитивных состояний. Например датчики и носимые устройства могут измерять частоту сердечных сокращений вариабельность сердечного ритма активность мозга и другие параметры связанные с уровнем концентрации и умственного напряжения. Компьютерное зрение способно отслеживать направление взгляда частоту миганий позу что косвенно свидетельствует об уровне внимания или утомления. В исследовании приведены результаты проведённого исследования психометрического тестирования с использованием валидизированных инструментов эмоциональной включённости в образовательный процесс и исследование факторов выгорания усталости от знаний студентов очной формы обучения Новороссийского филиала Финансового университета при Правительстве РФ. Выборка составила 125 респондентов.
Neuromelanin-sensitive magnetic resonance imaging (NM-MRI) enables non-invasive assessment of the substantia nigra pars compacta (SNpc). However, its utility in identifying dopaminergic dysfunction in idiopathic rapid eye movement sleep behavior disorder (iRBD) remains unclear. This study evaluated NM-MRI alterations in the SNpc, stratified by dopamine transporter (DaT) imaging status, to assess its performance in detecting early dopaminergic alterations. This combined retrospective and prospective study included 66 patients with iRBD (45 males, 67.0 ± 6.3 years) and 30 healthy controls (HCs) (11 males, 63.6 ± 6.5 years). Patients with iRBD were divided based on 18F-FP-CIT PET findings into those with abnormal (iRBD-CIT+) and normal (iRBD-CIT-) DaT imaging. NM metrics, including volume, signal-to-noise ratio (SNR), and contrast-to-noise ratio, were quantified using a template-based semi-automated method in the whole SNpc and its subregions (sensorimotor, associative, and limbic). Among the 66 patients with iRBD, 37 (56.1%) were classified as iRBD-CIT+. They exhibited significantly reduced NM metrics across the whole SNpc and all subregions compared to HCs. Notably, NM volume and SNR in the associative and limbic area were significantly lower in the iRBD-CIT+ group than in the iRBD-CIT- group, whereas other metrics did not differ. ROC analysis revealed fair performance for NM volume and SNR (AUC 0.73 and 0.75, respectively) in discriminating iRBD-CIT+ from iRBD-CIT-. Multivariable regression analyses confirmed these metrics as independent discriminators of DaT abnormalities. NM-MRI metrics, particularly limbic NM volume and SNR, serve as potential biomarkers for dopaminergic alterations in iRBD. These findings support NM-MRI as a strategic, radiation-free triage tool to identify high-risk patients requiring DaT imaging in iRBD.
Individuals with psychosis often experience levels of anxious avoidance comparable to patients with agoraphobia, causing detrimental effects on quality of life and functional outcomes. Automated virtual reality delivered cognitive behavioural therapy (CBT) may represent a promising treatment approach for addressing agoraphobic avoidance in patients with psychosis. We sought to evaluate the feasibility of using the virtual reality application gameChange as a treatment option for agoraphobic avoidance in people with psychosis in psychosis treatment units in Oslo, Norway. This study was conducted as a two arm non-randomized site-based feasibility trial (gameChange plus CBTp and standard care or gameChange plus standard care) across 3 treatment units for patients diagnosed with psychosis. Participants were allocated to intervention arm based on the type of CBTp available at their respective treatment units. It was aimed to recruit thirty patients with psychosis. Outcomes were assessed at baseline and post-treatment (13 weeks). Outcomes were recruitment-, retention-, and adherence-rates as well as patient satisfaction and a range of self-reported symptom measures. A total of 32 patients were referred: 28 patients were eligible, and 27 participated. 22 patients (81%) received an adherent dose of gameChange. 21 patients (77.7%) completed the post-treatment assessment. Participants reported high satisfaction with the intervention. gameChange was associated with improvements in symptomatology, most notably in the domains of functioning and depression. Findings suggest that incorporating virtual reality delivered cognitive behavioural therapy interventions in a Norwegian health care setting is feasible. Randomised controlled trials are warranted to assess treatment efficacy.