Accurate pneumonia segmentation is crucial for effective diagnosis and treatment of pneumonia, a major global health concern. While vision-based medical image segmentation models have shown promise, their heavy reliance on expert-level pixel annotations limits clinical use. Multi-modal segmentation methods aim to mitigate annotation needs by incorporating text guidance, yet face persistent challenges in optimally aligning high-resolution visual features with abstract textual features due to their inherent representational gap. To alleviate these issues, we introduce VLSCAM, a Vision-Language Sliding Cross Attention Model, which utilizes a newly proposed Sliding Cross Attention mechanism to achieve more rational utilization of both image and text features and more accurate pneumonia segmentation. Extensive experiments on the QaTa-COV19 and MosMedData+ datasets demonstrate that VLSCAM outperforms both uni-modal and multi-modal methods. Furthermore, even with only 5

Background and objective: To investigate changes in pulmonary vascular quantity and structure in patients with connective tissue disease (CTD) who have no interstitial lung abnormalities, to characterize quantitative pulmonary vascular differences in CTD without radiological ILD. Methods: This study included CTD patients who underwent chest high-resolution CT (HRCT) at our institution between January 2013 and June 2025. Radiologists excluded those with interstitial lung abnormalities, resulting in a cohort of CTD patients without interstitial lung disease (CTD-nonILD). Artificial intelligence–based quantitative CT software (AVIEW) was used to assess vascular parameters, including total vessel number (Ntotal), number of small vessels (<5 mm2, NCSA5), mean vessel diameter (VDmean), total vessel area (VAtotal), total lung blood volume (TBV), and blood volume of small vessels (BV5)—at 6, 9, 12, 15, and 18 mm from the pleura in the whole lung and individual lobes. Parameters were compared with healthy controls. Results: At the whole-lung level, Ntotal was significantly higher in CTD-nonILD patients at 6 mm and 9 mm depths (P < 0.05), with no significant differences at deeper zones (P > 0.05). VDmean, BV5, and BV5/TBV were significantly elevated in both whole lung and all lobes (P < 0.05). NCSA5 increased markedly at 6 mm depth but decreased at 12, 15, and 18 mm (P < 0.05). In upper lobes, NCSA5 was significantly higher at 6 mm (P < 0.05); no significant difference was seen in lower lobes (P > 0.05). Conclusion: CTD patients exhibit early pulmonary microvascular remodeling, characterized by peripheral proliferation of small vessels accompanied by relative small-vessel reduction in deeper parenchymal zones in CTD patients without radiological interstitial lung disease.

Currently, reliable preoperative methods for predicting vertebral artery (VA) invasion are lacking. The authors develop a novel model based on MRI radiomic signatures combined with clinical and imaging features for predicting intraoperative vertebral artery injury in patients with primary cervical tumors. Included in this retrospective study were 168 patients who received surgical resection for primary cervical tumors. They were randomly assigned to a training set (n = 117) and a test set (n = 51) . Least absolute shrinkage and selection operator logistic regression was applied for feature selection and radiomic signature construction. A multilayer perceptron (MLP) model and 10 machine learning models were used to develop diverse prediction models. Independent risk factors of clinical variables were screened by Logistic regression, based on which a clinical model was constructed. A combined model was established by combining the radiomic signatures and clinical factors. The predictive performance of the combined model was evaluated in both training and test sets using Hosmer–Lemeshow test and decision curve analysis (DCA). According to the scoring system, the MLP model obtained the highest total score of 87, meaning that its prediction performance was the best of all evaluated models, so the MLP was selected to construct the radiomics model. The AUC of the combined model in the training and test cohorts was 0.951 and 0.950 respectively, and both were higher than that of the radiomics model (AUC 0.900 in training set, p = 0.010, AUC 0.780 in test set, p = 0.001) and the clinical model (AUC 0.740 in training set, p < 0.001, AUC 0.781 in test set, p = 0.008) alone. The present study presents a nomogram that incorporates radiomic signatures and clinical features, which could be used to predict the risk of intraoperative VA injury in patients with primary cervical tumors.

Class imbalance in semi-supervised medical image segmentation poses a dual challenge: it not only compromises feature learning for tail classes but also introduces significant bias in loss gradients toward the predominant background class. To address these challenges, we introduce duo-component modulation network (DuoMod-Net), a synergistic learning framework integrating two specialized components. The first component, relative logarithmic modulation (RLM), addresses the dominant gradient bias by decoupling the background magnitude from foreground balancing. It establishes the background as a neutral pivot and then applies a relative, logarithmic scaling anchored by robust percentiles to preserve the dynamic range among the foreground organs. Concurrently, the second component, disagreement-driven adaptive feature refinement (DAFR), functions as a geometric regularization mechanism. It leverages intrinsic inter-model disagreement to selectively expand the feature space during training, forcing the decision boundary to recede. This expansion is removed at inference, establishing a safety margin that enhances detection reliability. Extensive validation across varying data regimes (5%, 10%, and 20%) demonstrates that DuoMod-Net yields substantial improvements on tail classes, increases detection reliability by minimizing catastrophic failures, and maintains robust zero-shot generalization on unseen datasets.

To develop and validate MRI-based radiomics models for predicting intraoperative massive blood loss (MBL) in patients with spinal metastases. A total of 507 patients diagnosed with spinal metastases were enrolled in this study, who were classified as MBL and non-MBL group, with the 2500 ml as the threshold. Radiomic features were extracted from T2WI and CET1 sequences and dimensionality reduction was performed by LASSO regression analysis. Radiomics models were developed using radiomics features, yielding a radiomics signature from the best model. Clinical variables were analyzed to create a clinical model. The combined model incorporated both clinical variables and Rad-signature. The predictive performance was assessed through the area under the curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1 score. Calibration curves and decision curve analyses (DCA) were generated to evaluate the model’s accuracy and clinical utility. Finally, a nomogram was developed to visualize the optimal model. Tumor vascularity, preoperative embolization, tumor location, and surgical segments were independent risk factors for MBL. Among 11 machine learning models evaluated, the AdaBoost model demonstrated optimal predictive performance (AUC = 0.775) and was selected as the radiomics model. The combined radiomics-clinical model achieved the highest discriminative ability (AUC = 0.818; 95

Deep-tissue visualization of Zn2+flux represents a significant yet challenging objective in molecular diagnostics[1].Since Zn2+is co-released with insulin during glucose-stimulated secretion,real-time imaging of its dynamics could provide a non-invasive window into β-cell function and the progression of diabetes[2].However,achieving this goal requires overcoming two major hurdles:sufficient tissue penetration and high detection sensitivity[3].In this context,the recent work by Bingbo Zhang and colleagues represents an important advance.They report a rationally designed Fe3+-based probe,FeL2,which in-tegrates a structurally reinforced Zn2+-responsive motif that effectively enhances relaxivity in deep-seated tissues such as the pancreas,offering a promising new tool for diabetes management and advancing the field of metabolic imaging[4].

Photodynamic therapy (PDT) faces critical challenges in practical application due to tumor hypoxia and the lack of precise imaging guidance. To address these limitations, we engineered Gd@CATCe6 (GCC) via catalase (CAT)-mediated biomimetic synthesis, where CAT serves as a structural template for the green synthesis of gadolinium-based nanoparticles, an enzymatic oxygenator through H2O2 decomposition, and a hydrophobic host for photosensitizer chlorin e6 (Ce6) loading. GCC leverages CAT's enzymatic activity to decompose tumor-overexpressed H2O2 into oxygen, effectively mitigating hypoxia while amplifying Ce6-mediated reactive oxygen species (ROS) generation under laser irradiation. In vitro studies confirmed a uniform nanostructure (approximately 10 nm), high longitudinal relaxivity (r1 = 10.9 mm-1s-1), and potent ROS production. In vivo magnetic resonance imaging (MRI) demonstrated significant tumor accumulation via the enhanced permeability and retention (EPR) effect, extending the imaging window to 1-2 h for precise therapy guidance. Notably, GCC combined with laser irradiation suppressed 4T1 tumor growth by 87.84% in mice, outperforming controls, while exhibiting good biocompatibility in blood and organ toxicity assays. This work presents an enzyme-based theranostic strategy that synergizes real-time imaging with self-oxygenating PDT, offering a promising solution to overcome hypoxia-driven therapeutic resistance.

Background:The growth of pulmonary nodules is an important predictor of malignancy. Their varied size makes them difficult to detect and monitor, and improper management may result in a high risk of malignancy. This study explores a non-invasive method for analyzing growth progression in pulmonary nodules between baseline and follow-up computed tomography (CT) scans. The approach integrates radiomics and deep features extracted from baseline imaging data. These features led to the development of a nodule growth prediction model to accurately differentiate growth status, thus providing an objective, quantitative basis for optimizing clinical follow-up strategies. Methods:We conducted a retrospective cohort study of patients with pulmonary nodules who underwent ≥2 chest CT examinations between 2021 and 2024, comprising 1,387 nodules (1,109 for internal training and 278 for internal testing) from baseline and follow-up CT scans. Nodules were classified into growth (volume increase ≥25%) and non-growth groups based on 1-year follow-up volumetric changes. From CT images, 2,264 radiomics features were extracted. Feature selection was conducted via least absolute shrinkage and selection operator (LASSO) regression, followed by principal component analysis (PCA) for dimensionality reduction. For deep feature extraction, we utilized a Swin Transformer architecture for nodule classification, extracting features from its penultimate layer which similarly underwent PCA-based dimensionality reduction. The extracted radiomics and deep features were integrated to develop logistic regression (LR) and random forest predictive models. Model performance was evaluated using the area under the curve (AUC), sensitivity, specificity, and additional metrics. Decision curve analysis (DCA) and calibration curves further evaluated clinical utility. Results:The combined model demonstrated strong predictive performance in predicting nodule growth in the Internal training set (n=1,109 nodules). It achieved an AUC of 0.902 [95% confidence interval (CI): 0.876-0.929], with a sensitivity of 0.815 and a specificity of 0.837, indicating a robust balance in diagnostic performance. The model significantly surpassed both radiomics-based (AUC =0.829; P<0.001) and deep learning-based (AUC =0.875; P<0.01) benchmarks. Furthermore, decision curve and calibration analyses consistently validated its clinical superiority and prediction accuracy. Conclusions:The integrated radiomics and deep learning model effectively stratifies pulmonary nodule growth risk. This approach provides clinicians with a quantitative tool for personalizing follow-up and intervention strategies, demonstrating significant potential for clinical translation.

In the quest to enhance medical consultation, our study introduces AI4Doctor, a sophisticated large-language model (LLM) tailored for the clinical domain. At the heart of AI4Doctor is an innovative integration strategy that synergizes distilled data extracted from electronic medical records (EMR) with empirical insights gathered from practicing physicians during the supervised fine-tuning. Although existing platforms offer informative responses, they fall short of replicating the nuanced decision-making processes of medical professionals, particularly in complex, integrative diagnostic scenarios. Motivated by the need to create a realistic medical practice environment, we propose that a combination of direct knowledge transfer from seasoned doctors and the strategic use of EMR can augment the abilities of LLM, enabling it to more closely mimic the clinical acumen of healthcare practitioners. To navigate the complexities of merging diverse instructional sources, we employ a curriculum learning approach during the fine-tuning process. Moreover, we advance our model’s performance by developing a reward system that incentivizes the alignment of the LLM’s outputs with the valuable attributes inherent in both doctors’ expertise, including diagnostic priors, risk thresholds, and heuristic saliencies accumulated from practice and EMR data. This is achieved through a novel reinforcement-learning approach. Besides, we introduce a new benchmark involving a comparative evaluation. We utilize a subjective evaluation system wherein experts critically assess the responses from a professional perspective as well. Our research underscores the potential of this hybrid model to serve as a robust tool in medical consultations, bridging the gap between artificial intelligence and real-world clinical practice.

BACKGROUND:Accurate assessment of smoke-inhalation-induced acute lung injury (SI-ALI) is essential for clinical diagnosis and effective management. Hyperpolarized 129Xe magnetic resonance (MR), an emerging imaging modality, has shown broad utility across various pulmonary conditions. This study aimed to evaluate the feasibility and potential of hyperpolarized 129Xe MR for assessing gas exchange impairment and microstructural alterations in the lungs following SI-ALI. METHOD:Two groups of rats (n = 5 per group) were studied. The smoke inhalation injury (SII) group was subjected to three 2-minute exposures to pine sawdust smoke under general anesthesia with endotracheal intubation. The sham group underwent identical procedures but was exposed to clean air. Twenty-four hours after exposure, pulmonary function tests, micro-computed tomography (micro-CT), and 129Xe MR examinations were conducted to obtain quantitative physiological parameters. Subsequently, lung tissues were harvested for histological analysis of alveolar septal wall thickness. RESULT:Rats exposed to smoke (SII group) showed significant decreases in lung volume, as indicated by reduced total lung capacity (TLC, p = 0.036) and forced vital capacity (FVC, p = 0.020). They also had signs of airway obstruction, with lower forced expiratory volume in 100 ms (FEV100, p = 0.041) and maximal mid-expiratory flow (MMEF, p = 0.014). Hyperpolarized 129Xe MR spectroscopy showed a lower red blood cell to tissue/plasma (RBC/TP) signal ratio in the SII group (0.45 ± 0.04) compared to the sham group (0.51 ± 0.04, p = 0.035), suggesting impaired gas exchange. The gas exchange time constant increased from 21.22 ms to 28.86 ms (p = 0.013), and the septal wall thickness measured by MR also increased (from 8.28 µm to 9.68 µm, p = 0.013). These MR results matched well with histological measurements, which also showed thickened alveolar walls (from 6.99 µm to 7.70 µm). Ventilation imaging revealed clear areas of reduced airflow, which corresponded to regions of lung consolidation seen on micro-CT scans. CONCLUSION:Hyperpolarized 129Xe MR enables quantitative evaluation of both functional and microstructural lung changes in a rat model of SI-ALI. These findings highlight its potential as a powerful noninvasive tool for assessing smoke-inhalation-induced acute lung injury.

CT emphysema segmentation supports objective COPD assessment, but pixel-level annotation is costly because emphysema lesions are spatially diffuse and visually heterogeneous. Existing lesion synthesis methods rarely model emphysema-specific low-attenuation parenchymal destruction, and synthetic-trained models often degrade on real CT scans. We propose a label-efficient emphysema segmentation framework built on two complementary mechanism-level components. First, Prior-guided Emphysema Synthesis (PES) generates synthetic lesions through Gaussian-based subregion selection, within-region density modulation, and prior-guided intensity sampling. Second, Restorative Contrastive Test-Time Training (ResCon-TTT) reduces the remaining domain gap by coupling multi-dimensional dynamic feature perturbation with a restoration-contrastive auxiliary objective. On the internal dataset, PES-trained UNet achieved 70.11% DSC, and ResCon-TTT further improved DSC to 73.42%. On two external datasets, ResCon-TTT achieved 72.63% and 83.35% DSC, outperforming competing TTT/TTA methods. These results suggest that emphysema-specific synthesis and feature-level test-time adaptation can reduce annotation dependence while improving cross-center robustness. The source code is publicly available at: https://github.com/smallrookie/ResCon-TTT.git.

Early diagnosis of knee osteoarthritis (KOA) remains challenging, particularly in distinguishing between Kellgren–Lawrence (KL) grades 1 and 2 on standard radiographs. This study aimed to develop a radiomics-based model using digital radiography (DR) to facilitate early identification of radiographic KOA (RKOA). A total of 859 patients with KL grade 1 or 2 were retrospectively enrolled and randomly divided into a training set (n = 601) and a validation set (n = 258). From anteroposterior and lateral DR images, 2,632 radiomics features were extracted per patient. Features were filtered using intraclass correlation coefficient (ICC ≥ 0.75), correlation analysis (threshold > 0.9), and least absolute shrinkage and selection operator (LASSO) regression, yielding 38 features. Five machine learning models were constructed and compared. Logistic regression (LR), which demonstrated the best generalizability, was used to compute a radiomics score (Radscore). A combined nomogram incorporating Radscore and age was developed and evaluated. In the validation set, the LR model achieved the highest area under the curve (AUC) of 0.821. The combined nomogram model outperformed the Radscore model alone, with AUCs of 0.914 vs. 0.908 in the training set (P = 0.0067), and 0.833 vs. 0.823 in the validation set (P = 0.0041). Calibration curves confirmed model goodness-of-fit, and decision curve analysis showed higher clinical net benefit for the nomogram. Digital radiography-based radiomics combined with age enables accurate early KOA diagnosis and demonstrates strong potential for clinical application.

Background:The white matter (WM) of patients who suffer from depression after long-term coronavirus disease 2019 (COVID-19) showed a dynamic change. However, to date, no research on the dynamic change in the WM of patients who develop severe depression after multiple COVID-19 infections has been conducted. This study aimed to evaluate long-term WM changes in patients who experience mild or moderate depression after their first COVID-19 infection followed by severe depression after their second infection. Methods:In total, 27 outpatients who developed severe depression after two COVID-19 infections (the patient group) and 28 outpatients who developed mild or moderate depression after one COVID-19 infection (the control group) were included in this prospective study. Psychological assessments for depression, anxiety, and insomnia were conducted 3-6 months after infection. Diffusion tensor imaging was performed. The fractional anisotropy (FA), mean diffusivity, axial diffusivity, and radial diffusivity were calculated and compared between the groups. The correlations between the psychological scores and image parameters were analyzed. Results:There was no difference between the two groups in terms of age (51.857±2.770 vs. 48.148±2.0678 years, P=0.222), gender (female: 75.0% vs. 55.6%, P=0.130), and education (P=0.317). The psychological scores were higher in the patient group than the control group (Patient Health Questionnaire-9: 17.290±2.600 vs. 8.640±2.599, Generalized Anxiety Disorder-7:15.440±3.490 vs. 7.640±2.376, Athens Insomnia Scale: 15.852±3.219 vs. 10.179±2.763; all P<0.001). The patient group showed decreased FA values in the bilateral corticospinal tracts, anterior thalamic radiations, and right superior longitudinal fasciculus (Pfamily-wise error <0.05). All the psychological scores were negatively mildly to moderately correlated with the FA values of the above-mentioned WM tracts (r=-0.491, r=-0.570, and r=-0.355, P<0.001, P<0.001, and P=0.008). Conclusions:The present study revealed impairments in the WM integrity of patients who developed severe depression within 3-6 months of repeated COVID-19 infections. The findings revealed the mechanism of COVID-19-related depression and underscored the importance of psychological intervention after COVID-19 infection.

ObjectivesTo evaluate the prediction value of radiomics models based on FDG-PET/CT for the therapeutic effect in patients with newly-diagnosed multiple myeloma (MM).Materials and methodsWe retrospectively reviewed the clinical characteristics and 18F-FDG-PET/CT imaging data of 165 MM patients. Randomly divided into a training set (n=133) and a test set (n=32) at a ratio of 8:2. All patients underwent whole-body PET-CT scans within one month prior to the commencement of treatment. Overall response rate was the principal efficacy endpoint, including stringent complete response (sCR), complete response (CR), very good partial response (VGPR), partial response (PR), disease stabilization (SD), and disease progression (PD). Deep response (DR) was defined as sCR, CR, and VGPR, while non-deep response included PR, SD and PD, 74 patients attained DR. Different models involving clinical, radiomics extracted from PET/CT, and their combination were constructed based on multiple logistic regression and logistic regression machine learning classifier after features selection, respectively. The models predicting performance were evaluated by the area under the ROC curve (AUC), sensitivity, specificity, accuracy, precision, and F1 score. Receiver Operating Characteristic (ROC) curves, decision curves, calibration curves, and DeLong’s test were applied to compare their ability.ResultsGender was the only one of clinical characteristics found to be independent prognosis factor for treatment evaluation, with a p-value of 0.041. The radiomics models outperformed the Clinical model significantly, among which the PET-CT model yielded the best results with the AUC of 0.809. The PET + CT + Clinical model achieved the optimal performance after integrating clinical and radiomic features, with the AUC of 0.813.ConclusionsThe FDG-PET/CT-based radiomics model, particularly when integrated with clinical features, can more effectively predict deep treatment response in newly diagnosed MM patients, offering significant clinical utility for early treatment stratification and personalized therapeutic guidance.

Pulmonary dynamic ventilation dysfunction is a common fea-ture of various lung diseases,including chronic obstructive pul-monary disease(COPD)[1],cystic fibrosis[2],and asthma[3].Regional assessment of ventilation dynamics offers substantial potential to enhance diagnostic accuracy and therapeutic monitor-ing in these conditions.Although current clinical evaluations pri-marily depend on global pulmonary function tests,emerging imaging modalities such as four-dimensional computed tomogra-phy(4D-CT)[4]and phase-resolved functional lung(PREFUL)imaging[5]enable temporal observation of structural and ventila-tion changes.However,these techniques are fundamentally lim-ited in their ability to visualize time-of-flight(TOF)gas flow patterns within airways and alveolar spaces-critical parameters for the direct assessment of regional gas diffusion efficiency.

Objective:This study aimed to construct and validate a fusion diagnostic model based on Fluorodeoxyglucose-Positron Emission Tomography/Computed Tomography(FDG-PET/CT) radiomics for predicting overall survival of multiple myeloma (MM) patients. Methods:A total of 199 patients newly diagnosed with MM were included from two centers. All patients underwent whole-body PET/CT scans within one month before the initiation of treatment and were followed up for over five years. Radiomic features of MM were extracted from CT images and dimensionality reduction was performed by LASSO regression analysis. Cox Proportional Hazards Model was then constructed to predict patient survival. A clinical-radiomic fusion model was constructed by integrating independent clinical risk factors, including comprehensive laboratory parameters, R-ISS, and PET functional metabolic parameters, with the radiomic model. The discrimination ability of the model was evaluated using the C-index, and it's calibration was assessed using calibration curves. Results:The C-indexes for the radiomics model in the training and testing cohorts were 0.736 and 0.708, respectively; for the clinical model, they were 0.676 and 0.696, respectively; and for the integrated model, they were 0.791 and 0.776, respectively. The integrated diagnostic model outperformed both the radiomics and clinical models, showcasing higher discriminative ability and improved calibration. In the training set, the C-index was 0.791 (95% confidence interval [CI]: 0.713-0.853), with an ICI of 0.015, E50 of 0.014, and AIC of 10.987. In the testing set, the C-index was 0.776 (95% CI: 0.654-0.894), with an ICI of 0.069, E50 of 0.04, and AIC of 11.492. Conclusions:This integrated prediction model exhibited satisfactory performance in predicting survival outcomes for patients diagnosed with MM and improved precision in discriminating between patients with a good prognosis and poor prognosis.

Emphysema, a diffuse and heterogeneous phenotype of chronic obstructive pulmonary disease (COPD), carries substantial morbidity and elevates lung cancer risk. While computed tomography (CT) aids in detection and monitoring, current deep learning methods depend on large annotated datasets. Unsupervised anomaly detection (UAD) provides an alternative but faces challenges with emphysema anomalies and weak emphysema semantics. In this study, we propose a self-supervised framework trained exclusively on non-emphysema CT scans using synthetically generated lesions to guide pixel-level anomaly modeling. We introduce EDLNet, an encoder-decoder architecture with spatial-channel refinement and adaptive feature fusion for emphysema detection and localization, followed by an unsupervised manner for emphysema staging. Multi-center evaluations show that our framework outperforms existing UAD approaches in detection and localization, while achieving a mean staging accuracy of 93.13% and a macro AUROC of 99.08%. This approach bridges clinical knowledge and artificial intelligence, offering a scalable and interpretable solution for lung disease analysis.

ObjectiveThis study aimed to evaluate whether deep learning-based image reconstruction (DLR) improves the accuracy of diffusion tensor imaging (DTI) measurements used to assess the severity of depression.MethodsA total of 52 patients diagnosed with depression in our hospital between March 2023 and July 2023 were enrolled in this study. The severity of depression was measured using the 9-item Patient Health Questionnaire (PHQ-9). Each patient underwent DTI scans. Two image sets were generated: one with the original DTI (ORI DTI) and one using DLR DTI. Tract-Based Spatial Statistics (TBSS) were used to compare the fractional anisotropy (FA) between DLR DTI and ORI DTI, as well as between patients with mild-to-moderate and those with severe depression. Multivariate logistic regression was carried out to determine independent factors for discriminating mild-to-moderate from severe depression patients. Receiver operating characteristic (ROC) curve analysis and areas under the curve (AUC) were used to assess the diagnostic performance.ResultsTwenty-eight patients with mild-to-moderate depression and 24 with severe depression were included. No significant differences were observed between the two groups in terms of gender (p = 0.115), age (p = 0.603), or educational background (p = 0.148). Compared to patients with mild-to-moderate depression, those with severe depression showed lower FA values in the right corticospinal tract (CST) on ORI DTI. Using DLR DTI, decreases in FA values were observed in the right CST, right anterior thalamic radiation, and left superior longitudinal fasciculus. The diagnostic model based on DLR DTI outperformed the ORI DTI model in assessing severity of depression (AUC: 0.951 vs. 0.764, p < 0.001).ConclusionDLR DTI demonstrated greater sensitivity in detecting white matter (WM) abnormalities in patients with severe depression and provided better diagnostic performance in evaluating severity of depression.

RATIONALE AND OBJECTIVES:To investigate the performance of two diagnostic models based on CT-derived lung and mediastinum radiomics nomograms for identifying cardiovascular disease (CVD) in Chronic Obstructive Pulmonary Disease (COPD) patients. MATERIALS AND METHODS:Hospitalized participants with COPD were retrospectively recruited between September 2015 and April 2023. Clinical data and visual coronary artery calcium score (CACS) were collected. Radiomics features of lung and mediastinum were extracted. Least absolute shrinkage and selection operator (LASSO) logistic regression was applied for feature selection and radiomic model construction. We constructed 3 radiomics models, based on lung, mediastinum, and combined lung-and-mediastinum. Multivariate logistic regression model was used to establish radiomics nomograms. The performance of radiomics nomograms was evaluated by area under the ROC curve (AUC) and decision curve analysis (DCA). RESULTS:Of 686 COPD patients, 131 had a history of CVD. Age, neutrophilic granulocyte percentage, hematocrit and GOLD stage were independent clinical factors for CVD. 12 lung, and 6 mediastinum radiomic features were collected to construct the radiomics models. As the lung-and-mediastinum radiomics model included the same 6 features as the mediastinum model, finally 2 radiomics models were studied (lung, mediastinum). The 2 radiomics nomograms showed better discriminatory ability (AUC: 0.79, 95%CI [0.72, 0.86] for lung; 0.86, 95%CI [0.81, 0.92]) for mediastinum) than the clinical factors model (AUC: 0.71, 95%CI [0.64, 0.78]) and visual CACS (AUC: 0.65, 95%CI [0.57, 0.72]). DCA demonstrated the 2 radiomics nomograms outperformed the clinical factors and CACS across the majority of the range of reasonable threshold probabilities. CONCLUSION:We developed chest CT-based nomograms to identify CVD in COPD patients, in particular based on mediastinum features, had better discriminatory power than clinical factors and visual CACS. TRIAL REGISTRATION:This retrospective study was approved by the institutional review boards at Second Affiliated Hospital of Naval Medical University, Tongji Hospital of Tongji University and Sir Run Run Shaw Hospital (ChiCTR2300069929 March 29, 2023). Retrospectively registered.

Purpose To develop a deep learning model that uses a single inspiratory chest CT scan to perform parametric response mapping (PRM) and predict functional small airways disease (fSAD). Materials and Methods In this retrospective study, predictive and generative deep learning models for PRM using inspiratory chest CT were developed using a model development dataset with fivefold cross-validation, with PRM derived from paired respiratory CT as the reference standard. Voxelwise metrics, including sensitivity, area under the receiver operating characteristic curve (AUC), and structural similarity index measure, were used to evaluate model performance in predicting PRM and generating expiratory CT images. The best-performing model was tested on three internal test sets and an external test set. Results The model development dataset of 308 individuals (median age, 67 years [IQR: 62-70 years]; 113 female) was divided into the training set (n = 216), the internal validation set (n = 31), and the first internal test set (n = 61). The generative model outperformed the predictive model in detecting fSAD (sensitivity, 86.3% vs 38.9%; AUC, 0.86 vs 0.70). The generative model performed well in the second internal (AUCs of 0.64, 0.84, and 0.97 for emphysema, fSAD, and normal lung tissue, respectively), the third internal (AUCs of 0.63, 0.83, and 0.97), and the external (AUCs of 0.58, 0.85, and 0.94) test sets. Notably, the model exhibited exceptional performance in the preserved ratio impaired spirometry group of the fourth internal test set (AUCs of 0.62, 0.88, and 0.96). Conclusion The proposed generative model, using a single inspiratory CT scan, outperformed existing algorithms in PRM evaluation and achieved comparable results to paired respiratory CT. Keywords: CT, Lung, Chronic Obstructive Pulmonary Disease, Diagnosis, Reconstruction Algorithms, Deep Learning, Parametric Response Mapping, X-ray Computed Tomography, Small Airways Supplemental material is available for this article. © The Author(s) 2025. Published by the Radiological Society of North America under a CC BY 4.0 license. See also the commentary by Hathaway and Singh in this issue.