Current Medical Imaging Reviews最新文献

Objective: This study aimed to extract radiomic features from ultrasound (US) images of soft tissue tumors (STTs) and develop a diagnostic model for STTs using radiomic and clinical patient data.

Methods: Three hundred and sixty-nine patients were recruited as the training group, with 249 benign and 120 malignant STTs, and 127 patients as the validation group, with 93 benign and 34 malignant STTs. We extracted the radiomic features of the US images using an open-source Python package. We selected the most relevant features using the least absolute shrinkage and selection operator (LASSO) regression. Then we used a combination of clinical indexes, radiomic features, and color-Doppler US to construct a diagnostic model for STTs. The diagnostic performance of the model was evaluated by measuring its sensitivity, specificity, area under the receiver operating curve (AUC), and calibration.

Results: We selected 20 radiomic features of the US images. The model based on the clinical indexes, radiomic features, and color-Doppler scores showed good diagnostic performances on both the training [AUC: 0.97 (0.95-0.98)] and validation datasets [AUC: 0.93 (0.86-0.99)]. The model also presented good calibration with the original results.

Conclusion: The diagnostic model based on clinical, US radiomic, and imaging features presented a high diagnostic performance in STTs, which can have potential value in further clinical utilization.

Objective: This study aimed to explore the feasibility of habitat radiomics based on Non-Contrast Computed Tomography (NCCT) for differentiating Pleomorphic Adenoma (PA) and Adenolymphoma (AL), and to compare it with both clinical and conventional radiomics models.

Methods: A retrospective collection of clinical and imaging data was conducted on 203 patients who underwent pathology-proven procedures from October 2015 to August 2024 at two hospitals. Tumor Regions of Interest (ROIs) were delineated on NCCT images, and the K-means algorithm was used to jointly cluster the training and validation sets. Radiomics features were extracted, followed by feature selection using the Minimal-Redundancy- Maximal-Relevance (mRMR) and Least Absolute Shrinkage and Selection Operator (LASSO) methods. Univariate and multivariate logistic regression analyses were conducted to identify clinical independent risk factors. The clinical, radiomics, and habitat models were constructed after selection of the clinical and radiomics features. The optimal radiomics model was combined with independent clinical risk factors to develop a nomogram and a combined diagnostic model. The performance of each model was evaluated using the Area Under the Receiver Operating Characteristic (ROC) Curve (AUC), and the DeLong test was used to compare model performance. Calibration curves and Decision Curve Analysis (DCA) were utilized to evaluate model calibration and clinical net benefit, respectively.

Results: Four distinct habitat areas were identified through clustering analysis. The habitat_all model achieved superior predictive performance, with AUCs of 0.903 in the training set and 0.846 in the validation set. This model outperformed the clinical model (training set AUC: 0.837; validation set AUC: 0.823), the conventional intra-tumor radiomics model (training set AUC: 0.845; validation set AUC: 0.840), and each of the four individual habitat models (training set AUCs: Habitat1 = 0.839, Habitat2 = 0.847, Habitat3 = 0.822, Habitat4 = 0.859; validation set AUCs: Habitat1 = 0.823, Habitat2 = 0.840, Habitat3 = 0.827, Habitat4 = 0.842). Furthermore, the nomogram integrating clinical independent risk factors (age and smoking history) with the habitat_all model showed improved predictive performance (AUCs for the training and validation sets were 0.953 and 0.883, respectively) and demonstrated significant clinical net benefit.

Conclusion: Habitat radiomics analysis based on NCCT enables accurate differentiation between PA and AL, providing novel insights for clinical diagnosis and treatment.

Introduction: This review aims to evaluate the accuracy of Magnetic Resonance Imaging (MRI) in diagnosing Non-Alcoholic Fatty Liver Disease (NAFLD) based on the published literature.

Methods: A PubMed search was performed using the keywords NAFLD and MRI, and the literature search deadline was set before April 2025.

Results: A total of 86 studies out of 405 retrieved were included in this study. The results showed that Magnetic Resonance Imaging Proton Density Fat Fraction (MRI-PDFF) was positively correlated with steatosis grading. The proton Density Fat Fraction Magnetic Resonance Spectroscopy (PDFFMRS) threshold of 5% could be used to diagnose liver steatosis. Apparent Diffusion Coefficient (ADC) of NAFLD patients was significantly lower than that of controls. A 15% increase in Magnetic Resonance Elastography (MRE) was the strongest predictor of progression to advanced fibrosis in NAFLD. The corrected T1 (cT1) cutoff value of 875 ms was used to identify liver fibrosis in NAFLD. The correlation between the Liver Surface Nodules (LSN) score and the stage of fibrosis in NAFLD was very strong. Dynamic enhanced MRI (DCE-MRI) parameters increased with increasing severity of NAFLD and fibrosis.

Discussion: This study evaluated the value of multiple MRI techniques in diagnosing NAFLD, confirming MRI's high accuracy and reliability as a noninvasive tool for quantifying NAFLD. However, future technical specification harmonization is needed to enhance comparability of results and validate generalizability through multicenter studies.

Conclusion: MRI is a highly reliable and accurate method for diagnosing NAFLD.

Background: Disseminated tuberculosis (dTB) can occur in immunocompetent adults, frequently mimicking metastatic malignancy, thereby delaying the diagnosis.

Case presentation: A young woman without known immunosuppression developed multisystem disease involving the peritoneum/ovaries, hepatobiliary structures, lymph nodes, adrenals, and thoracolumbar spine. CT/MRI and PET/CT suggested widespread neoplastic disease. Because FDG avidity is nonspecific, we prioritized histologic confirmation. Surgical exploration and targeted biopsies showed necrotizing granulomatous inflammation compatible with tuberculosis; microbiologic testing supported the diagnosis. The patient commenced directly observed first-line therapy (isoniazid, rifampin, pyrazinamide, ethambutol) as the intensive phase, followed by an isoniazid-rifampin continuation phase. Under treatment, symptoms improved, and interval imaging showed regression of inflammatory lesions.

Conclusion: In cancer-like, multisystem presentations, even in apparently immunocompetent hosts, tissue diagnosis is decisive, and imaging should primarily guide sampling. Early recognition and standardized therapy can prevent irreversible morbidity.

Introduction: This study aimed to evaluate the predictive value of radiomic features derived from 18F-FluoroDeoxyGlucose (FDG) PET/CT for Epidermal Growth Factor Receptor (EGFR) gene mutations in patients with lung adenocarcinoma.

Methods: A retrospective analysis was conducted on 93 patients diagnosed with solitary lung adenocarcinoma who underwent 18F-FDG PET/ CT imaging and EGFR mutation results. The patients were divided into training (46 cases) and testing (47 cases) cohorts. Radiomic features were extracted from the primary tumor sites' PET and CT images. Feature selection was performed using the Mann-Whitney U test and least absolute shrinkage and selection operator (LASSO) regression. A radiomics score (Rad-score) was constructed, and combined models incorporating clinical factors and metabolic parameters were developed. Predictive performance was evaluated using receiver operating characteristic (ROC) curves, area under the curve (AUC), accuracy, and decision curve analysis (DCA).

Results: The radiomics model achieved AUCs of 0.865 (95% CI: 0.747-0.983) and 0.737 (95% CI: 0.572-0.901) in the training and testing sets, respectively, with corresponding accuracies of 80.9% and 78.3%. The clinical model alone demonstrated inferior performance, with AUCs of 0.637 and 0.645. The combined model showed slightly improved AUCs (0.885 and 0.714) but did not significantly outperform the radiomics-only model (P > 0.05). DCA indicated greater clinical utility for the radiomics model across a wide range of threshold probabilities.

Discussion: PET/CT-based radiomics research has also achieved good efficacy in predicting EGFR gene mutations. Compared with morphological imaging techniques, such as X-ray, ultrasound, and CT, 18F-FDG PET/CT imaging has the significant advantage of providing functional and metabolic information of lesions. Both radiomics and composite models could predict EGFR mutation status in lung adenocarcinoma patients, but the radiomics model showed slightly better clinical predictive efficacy than the composite model.

Conclusion: The radiomics model and the combined model integrating Rad-score with clinical factors demonstrated comparable abilities in effectively predicting EGFR mutation status in patients with lung adenocarcinoma. These models could offer a non-invasive approach for identifying EGFR mutations.

Introduction: This study aimed to predict the occurrence of cardiac remodeling and/or myocardial fibrosis using machine learning based on T1 mapping from cardiovascular magnetic resonance in athletes.

Methods: A total of 104 athletes and 20 healthy sedentary controls underwent a 3.0T cardiovascular magnetic resonance scan. Cardiac function parameters, T1 and extracellular volume, were measured for 16 segments of the left ventricle. These parameters were separately compared between athletes and controls, and between the positive and negative athlete groups. Gradient boosting machines, logistic regression, classification and regression trees, and support vector machines were constructed for the prediction of cardiac remodeling and/or myocardial fibrosis.

Result: Higher extracellular volume values of segments 1,4,6,8, and 9 and lower native T1 values of segments 8 and 14 were found in athletes than controls (p<0.05). Native T1 values of segments 3,6,8,9,10,14, and 15 and extracellular volume values of segments 3,6, and 8 were higher in the positive athletes group than those in the negative athletes group (p<0.05). The most effective model was the Gradient Boosting Machine, with an AUC of 0.899, an accuracy of 82.7%, a sensitivity of 90.0%, and a specificity of 81.0%. The top three important factors were: the native T1 value of segment 10, the extracellular volume value of segment 3, and body surface area.

Conclusion: Native T1 and extracellular volume values increased in athletes with cardiac remodeling, which may reveal the relationship between cardiac remodeling and myocardial fibrosis. Early cardiac magnetic resonance imaging is performed to monitor athletes' native myocardial T1 and ECV values, assess their risk levels, and guide subsequent surge planning to reduce the risk of adverse cardiovascular events. A GBM model with better performance can predict adverse cardiovascular events based on T1 mapping parameters, and the prediction can be verified by tracking the subsequent athlete's status.

Introduction: Pseudomyogenic Hemangioendothelioma (PMHE), also known as epithelioid sarcoma-like hemangioendothelioma, is a rare, indolent, low-grade vascular tumor. It typically presents as firm cutaneous nodules, with a predilection for the lower extremities and a male predominance. While numerous cases have been reported in pathology literature, detailed radiologic descriptions, particularly of soft tissue origins, are scarce. This report aims to bridge this gap by presenting a rare case of PMHE with comprehensive imaging findings.

Case presentation: We report on a 67-year-old male who presented with painful, palpable papules on his right buttock. MRI revealed multifocal dermal nodules demonstrating low signal intensity on T1-weighted images and high signal intensity with a distinctive peripheral high-signal halo on T2-weighted images. Notably, T1 gadolinium fat-saturated sequences exhibited marked enhancement with a characteristic peripheral rim enhancement pattern. The lesions were confined to the cutaneous layer. Initial radiological differentials included post-inflammatory granuloma and sarcoma. Histopathological examination confirmed PMHE. PET/CT demonstrated no evidence of systemic metastasis, and the patient has remained recurrence-free for two years following surgery.

Conclusion: This report highlights a rare case of cutaneous PMHE and details its distinctive MRI features, particularly the peripheral rim enhancement. Given its rarity and often non-specific clinical and imaging presentations, there is a significant potential for misdiagnosis. Therefore, it is crucial for radiologists to be aware of PMHE and familiarize themselves with its characteristic radiological patterns to facilitate accurate, timely diagnosis and ensure appropriate patient management.

Background: Non-invasive biomarkers of liver metabolism are essential for early detection of metabolic alterations. Choline plays a central role in hepatic function, yet its dietary intake and imaging correlates remain underexplored. This study evaluated the feasibility of proton Magnetic Resonance Spectroscopy (1H-MRS) at 3T for hepatic choline quantification and examined its correlation with dietary intake in young women, a population at risk of nutrient-sensitive liver conditions.

Methods: In this prospective cohort study, 88 healthy female radiology students (mean age: 21.4 ± 1.8 years) underwent single-voxel 1H-MRS of the liver using a 3T Siemens Magnetom Vida scanner. Spectra were acquired with a point-resolved spectroscopy (PRESS) sequence (TR = 2000 ms, TE = 40 ms, voxel size = 20 × 20 × 20 mm3), with automated shimming and unsuppressed water referencing. Spectral analysis was performed using LCModel (v6.3), applying quality thresholds (Signal-to-Noise Ratio (SNR) > 5, linewidth < 0.1 ppm, Cramér–Rao Lower Bound (CRLB) < 20%. Hepatic choline concentrations were expressed in Institutional Units (IU). Dietary intake was assessed using a validated Food Frequency Questionnaire (FFQ).

Results: High-quality spectra were consistently obtained (mean SNR: 12.6 ± 3.1; linewidth: 0.048 ± 0.012 ppm). Mean hepatic choline concentration was 4.63 ± 1.21 IU, while mean dietary intake was 29.1 ± 8.7 mg/day. A significant positive correlation was observed (r = 0.555, p < 0.001). Regression analysis confirmed dietary intake as a significant predictor (β = 0.56, R2 = 0.308, p < 0.001).

Discussion: These findings demonstrate that ¹H MRS at 3T provides reproducible hepatic choline quantification and captures meaningful variability linked to dietary intake. The observed correlation highlights the potential of MRS as a translational biomarker of nutrient related liver metabolism. Integrating MRS into multiparametric liver imaging protocols may enhance early detection of metabolic alterations and broaden the scope of noninvasive liver assessment.

Conclusion: 1H-MRS at 3T is a feasible and reproducible technique for hepatic choline quantification. By measuring metabolites directly in the liver at their site of production, rather than in circulation, where concentrations may be altered, MRS provides physiologically relevant insights into nutrient-related hepatic metabolism. Its correlation with dietary intake highlights its potential as a translational imaging biomarker for early detection and risk stratification of nutrient-sensitive liver conditions.

Introduction: Kidney abnormalities such as cysts, stones, tumors, and other structural disorders pose significant health risks and can lead to chronic kidney disease if not diagnosed in time.

Materials and methods: This study proposes a deep learning-based diagnostic framework that introduces an enhanced feature extraction strategy through a novel model known as Kidney Transformer Network (KTNET). The system is designed to automatically detect and classify multiple kidney conditions by effectively extracting disease-specific features from CT scan images. By leveraging transformer-based architecture, KTNET improves feature representation and enables highly accurate discrimination between Normal, Cyst, Tumor, and Stone cases.

Results: Experimental results demonstrate that the proposed model achieves outstanding diagnostic performance, recording 99.7% accuracy, 99.4% precision, 99.3% recall, and a 99.6% F1-score, surpassing traditional image processing methods and several existing deep learning models.

Discussion: The model's adaptability and efficiency with diverse CT scan images highlight its potential for practical integration in clinical workflows.

Conclusion: This research advances medical imaging by providing an intelligent, reliable, and accurate framework for the early detection and classification of kidney abnormalities, ultimately enhancing patient diagnosis and clinical decision-making.