Current Medical Imaging Reviews最新文献

Objective: This study aimed to develop and validate a deep learning radiomics (DLR) model based on superb microvascular imaging (SMI) for the noninvasive assessment of the severity of arteriolosclerosis in patients with chronic kidney disease (CKD).

Materials and methods: From June 2022 to December 2024, we prospectively recruited 326 CKD patients who underwent kidney biopsy across two medical centers. The enrolled patients were randomly allocated to the training or testing set in a 7:3 ratio. Deep learning (DL) features and radiomics features from SMI images were extracted, and after dimensionality reduction, they were used to establish deep learning radiomics (DLR) models. The performance of the proposed models was assessed through receiver operating characteristic (ROC) analysis and decision curve analysis (DCA).

Results: Among the 326 CKD patients, 165 were positive for arteriolosclerosis and 161 were negative. In the training group, the area under the curve (AUC) values for the CDUS model,clinical model, radiomics model, DL model, and DLR model were 0.621 (0.547-0.695), 0.68 (0.611-0.749), 0.763 (0.703-0.823), 0.820 (0.767-0.874), and 0.840 (0.790-0.890), respectively. In the testing group, the AUCs were 0.677 (0.571-0.783), 0.776 (0.684-0.869), 0.727 (0.626-0.829), 0.779 (0.687-0.872), and 0.819 (0.735-0.903), respectively. The DLR model outperformed standalone radiomics, DL models, and the CDUS-based clinical model. The DCA validated the clinical utility of the DLR model.

Conclusion: The DLR model utilizing SMI imaging can precisely and non-invasively assess the severity of arteriolosclerosis in CKD patients, which can assist physicians in formulating more favorable treatment plans for patients.

Objective: This study aimed to investigate the clinical and MR imaging differences between dual-phenotype hepatocellular carcinoma (DPHCC) and conventional hepatocellular carcinoma (HCC).

Methods: A retrospective analysis was conducted on the clinical data and MRI findings of 29 patients with DPHCC and 29 propensity score-matched patients with conventional HCC, confirmed by surgical pathology, from January 2019 to January 2022 at Fudan University Zhongshan Hospital. Clinical characteristics, lesion location, morphology, size, signal intensity, enhancement patterns, vascular invasion, and lymph node metastasis were analyzed for both groups.

Results: Between the DPHCC group and the HCC group, statistically significant differences were found in cirrhosis, pathological grade, lesion morphology, enhancement patterns, delayed capsular enhancement, and lymph node metastasis. There were no statistically significant differences between the two groups in terms of age, gender, hepatitis B infection, AFP, CA199, microvascular invasion (MVI), capsular invasion, lesion size, location, vascular invasion, ADC values, and T1WI and T2WI signals.

Conclusion: Compared to HCC, DPHCC has a higher pathological grade, more irregular lesion morphology, and a higher incidence of both fast-in and slow-out and slow-in and slow-out enhancement patterns, as well as higher rates of lymph node metastasis. The findings have provided valuable insights for the accurate diagnosis of DPHCC.

Introduction: This study introduces a novel fusion of 3D ResNet classification and Grad-CAM visualization to predict vaginal invasion in early-stage cervical cancer using T2WI-MRI, enhancing diagnostic accuracy while enabling anatomical localization of invasive lesions.

Methods: This retrospective study analyzed sagittal T2WI from 160 patients with pathologically confirmed stage IB-IIA cervical cancer to predict vaginal invasion. Following an 8:2 training-test split, radiomic features were extracted from manually delineated intratumoral regions and four concentrically expanded peritumoral regions (1-4mm). Features selection by Pearson correlation and LASSO regression. Random forest models incorporating intratumoral and peritumoral (0-4mm) features were constructed, with ROC analysis identifying the optimal model. Subsequently, a 3D-ResNet architecture, enhanced with anisotropic convolutional layers and sophisticated data augmentation, was developed and optimized using the optimal ROI configuration. Model interpretability was facilitated using Grad-CAM, with performance assessed by AUC, sensitivity, specificity, accuracy, and precision.

Results: The AIC-enhanced 3D ResNet-18 model, integrating intratumoral and 3mm peritumoral regions, showed superior test performance (AUC: 0.784, Sensitivity: 0.650, Specificity: 0.765, Accuracy: 0.611, Precision: 0.686) versus the baseline (AUC: 0.742), representing a 6% AUC improvement. Grad-CAM heatmaps identified diagnostically relevant regions within the tumor microenvironment, enhancing biological plausibility and model interpretability.

Discussion: This attention-integrated 3D ResNet-18 framework (AUC=0.784) facilitates non-invasive vaginal invasion detection for fertility-sparing decisions, validated through Grad-CAM tumor localization; however, derivation from a single-center cohort warrants external validation and prospective studies before clinical translation.

Conclusion: This preliminary study demonstrates promising deep learning performance (3D ResNet-18+Grad-CAM+AIC) for vaginal invasion assessment, despite moderate n; however, a single-center retrospective design limits generalizability.

Introduction: Pneumonia is a serious respiratory disease that requires early and precise diagnosis to reduce morbidity and mortality. This study aims to develop an efficient deep learning model for the accurate classification of pneumonia, COVID-19, and normal cases using chest X-ray and CT images.

Methods: The proposed model combines Capuchin Red Kite Optimization (CRKO) with a Swin Transformer-based Convolutional Block Attention Module (ST-CBAM). A Butterworth filter is applied during preprocessing to enhance image quality. ResNet and Vision Transformer are used for feature extraction, capturing local and global patterns, respectively. These features are fused using Adaptive Gated Recurrent Units (AGRU) and optimized with CRKO. The model is trained and validated using a publicly available chest X-ray dataset from Kaggle.

Results: The model achieved classification accuracies of 99% for normal, 99.9% for COVID-19, and 98.2% for pneumonia cases. It recorded an AUC of 98.93%, outperforming existing models such as ACNN, 3D-CNN, LWHNN, and CA-DCNN in both accuracy and execution time.

Discussion: The integration of CRKO with ST-CBAM, along with hybrid feature extraction and fusion techniques, contributes to the model's high performance. The results indicate a strong potential for clinical application. However, future studies should validate the model across diverse, realworld datasets to ensure generalizability.

Conclusion: The proposed deep learning framework offers a fast, accurate, and reliable solution for automated pneumonia diagnosis, showing promise for deployment in medical imaging systems.

Introduction: Life-threatening respiratory conditions such as COVID-19 and pneumonia demand rapid and accurate diagnosis. Chest X-rays (CXR) are widely used due to their accessibility and cost-effectiveness, but interpreting them remains clinically challenging, especially with overlapping radiological features.

Methods: The proposed VSAG-HDL Net, a novel hybrid deep learning framework designed to enhance the accuracy and interpretability of CXR-based diagnosis. The architecture integrates a Variational Spatial Attention Fusion U-Net (VSA-FU-Net) for lesion segmentation and a Sharpened Cosine Similarity (SCS) Network for disease classification. A dataset of 21,165 CXR images from the Radiography Database was used. Segmentation performance was evaluated using Dice Similarity Coefficient (DSC) and Intersection over Union (IoU), while classification performance was assessed via accuracy metrics.

Results: The VSA-FU-Net achieved a DSC of 90% and an IoU of 95%, indicating high precision in localizing lesions of varying shapes and sizes. The classification module reached an overall accuracy of 95.5%, outperforming traditional CNN-based methods such as CoroDet (+4.3%), CovXNet (+5.3%), and ShuffleNet (+3.9%). Although slightly less accurate than DenseNet+VIT (-2.0%) and DenseNet+VIT+GAP (-2.3%), the proposed framework offers competitive accuracy with significantly reduced model complexity.

Discussion: The elimination of redundant feature extraction and the integration of spatial attention enhance both the diagnostic performance and computational efficiency, making the framework suitable for real-world clinical settings.

Conclusion: VSAG-HDL Net provides a robust, interpretable, and resource-efficient solution for chest disease detection in CXR. Its clinical integration can support early and accurate diagnostic decision-making, particularly in resource-limited environments.

We aim to enhance the accuracy of healthcare AI by generating realistic synthetic medical images using Exposed GANs. One potential issue with synthetic MIG using exposed GANs is that the generated images may not accurately reflect real medical images, which could lead to incorrect medical AI diagnostic decisions. The primary goal of this research is to examine the capacity of GANs for generating synthetic medical images, which can improve the accuracy of healthcare AI systems. It is preferable to collaborate with medical institutions or utilize publicly available datasets from the Medical Segmentation Decathlon (MSD) to obtain medical images for academic research. One well-known pre-processing method for medical image data is normalizing to ensure all pixel values fall within a certain range. In the meantime, the Exposed GAN architecture has been designed to incorporate adversarial aerial training, aiming to produce more realistic medical images by pitting the generator against the discriminator to enhance output quality while improving the discriminator's ability to distinguish between fake and real images. Customization is a more likely research strategy; one can optimize model input parameters and loss functions (or offset the increased computational task of acquiring conditional GANs) at the architecture level. Data augmentation techniques, including random transformations and domain-specific adjustments, are employed to leverage the integration of synthetic data models and enhance the realism and generalization capabilities of the generated images. To enhance the accuracy of healthcare AI using synthetic MIG with exposed GANs, Python code must be implemented to train the GAN model on medical image datasets. The output performances of the discriminator were as follows: discriminator accuracy was 0.6924 on the real data and 0.78789 on the fake data. The average accuracy rate, MPa, was 96.29%, which serves as an evaluation tool for the success of our single-generator GAN in encouraging fabrication applications. There is intense hope that we will be able to unify synthetic MIG-GAN techniques to promote other health AI algorithms for personal applications.

Introduction: Synchronous liver metastases (SLM) critically influence prognosis in rectal cancer, highlighting the need for accurate preoperative detection. This study aimed to compare the predictive performance of radiomic features extracted from T2-weighted imaging (T2WI) and diffusion-weighted imaging (DWI) MRI sequences and to develop machine learning-based predictive models for the early detection of SLM in rectal cancer patients.

Methods: This retrospective study included 160 rectal cancer patients confirmed by pathology at our institution between September 2018 and June 2023. After screening, 137 patients were enrolled, comprising 71 patients with SLM and 66 without SLM. Clinical characteristics such as age, gender, tumor (mrT) staging, lymph node (mrN) staging, tumor size, tumor distance from the anal verge, location, and circumferential range were analyzed, with mrT and mrN staging showing statistical significance (p < 0.012). Radiomic features were extracted from regions of interest (ROIs) on T2WI and DWI using Pyradiomics after manual segmentation in ITK-SNAP. A total of 3,452 radiomic features (1,726 each from T2WI and DWI) were extracted, of which 14 features (4 from T2WI and 10 from DWI) were selected using the LASSO. Predictive models were developed using three machine learning algorithms: Logistic Regression (LR), Support Vector Machine (SVM), and Random Forest (RF), with a five-fold cross-validation strategy.

Results: Among the machine learning algorithms, the RF consistently outperformed LR and SVM across all models. The Optimal model yielded the highest predictive performance, with RF achieving an AUC of 0.82 (95% CI: 0.66-0.93), an accuracy of 0.71, and an F1-score of 0.74. RF also showed superior performance in the Combined-Optimal model (AUC = 0.76, accuracy = 0.71). In contrast, models built using LR and SVM algorithms demonstrated moderate performance, with lower AUC values ranging from 0.68 to 0.70. Confusion matrix analysis confirmed RF's superior classification ability, accurately predicting SLM and non-SLM cases.

Discussion: The incorporation of radiomics and RF-based models conveys a promising, non-invasive approach for enhancing early detection and risk stratification of SLM, which could help with more reliable clinical decision-making and individualized treatment planning for patients with rectal cancer.

Conclusion: The optimal feature set-based predictive model demonstrated the highest accuracy for SLM detection, with the RF algorithm outperforming LR and SVM by consistently achieving the best AUC and balanced diagnostic performance.

Introduction: Transanal small bowel evisceration is an extremely rare and life-threatening surgical emergency that primarily occurs in debilitated elderly patients. Preoperative computed tomography (CT) can be useful for identifying the viability of eviscerated small bowel and other intra-abdominal pathologies.

Case presentation: In this study, we report the case of an 81-year-old woman who presented with sudden anal protrusion of small bowel loops. Computed tomography (CT) demonstrated a rectal wall defect, pneumoperitoneum, and herniation of the small bowel with features suggestive of strangulation. Emergency laparotomy revealed a firmly impacted ileal segment plugging a perforation at the rectosigmoid junction, likely due to increased intraabdominal pressure, necessitating small bowel resection and the Hartmann procedure. Early diagnosis and prompt surgical intervention led to a favorable postoperative course.

Conclusion: This case highlights the critical role of CT in identifying rectal perforation and intrarectal small bowel evisceration.

Introduction: This study aimed to systematically review the application of narrow band imaging (NBI) in the diagnosis, treatment, and follow-up of head and neck cancer.

Methods: Through literature review and generalization of our clinical experiences, this review thoroughly described the features, mechanisms, advantages, drawbacks, and prospects of NBI in the treatment of head and neck cancer.

Results: NBI is an emerging endoscopic technology that emits an ambient light at wavelengths of 415 nm (blue) and 540 nm (green) to clearly visualize the details on the mucosal surface. It presents potent efficiencies in the preoperative, intraoperative, and postoperative surveillance and diagnosis of head and neck cancer.

Conclusion: NBI is a front-edge imaging technology that allows early screening, precise treatment, and postoperative monitoring of head and neck cancer.

Background: Although airway size changes occur in patients with chronic asthma, HRCT has not yet been used to assess changes in the inner diameter of the airways.

Objective: This study aimed to evaluate the airway diameter in asthma recovery by using HRCT.

Methods: Thirty patients with asthma were recruited and underwent HRCT examination in acute exacerbation and stable phase, respectively. The inner diameter of the airway (Din) was measured from the bilateral main bronchi to all 18 segmental bronchi during acute exacerbation and the stable phase.

Results: The inner diameter of the airway reduced significantly in the acute exacerbation period compared to the stable period (P < 0.05). The mean Din reduction (%) in segmental bronchi was 12% and that in lobar bronchi was 6%. Among the 30 patients, the dorsal segmental bronchi of both lower lobes showed the highest incidence of stenosis during acute exacerbation compared to the stable phase (right: 18 cases; left: 16 cases), while the lingular bronchus exhibited the highest stenosis incidence among lobar bronchi (18 cases). Although the number of stenotic segmental and lobar bronchi demonstrated a positive correlation with disease severity across mild, moderate, and severe groups, no statistically significant differences were observed in intergroup comparisons (P>0.05).

Conclusion: CT images of bronchial stenosis showed obvious dilation after appropriate medication, and the inner diameter of the airway can be used as a practical and convenient index to evaluate the recovery of asthma.