This study aims to predict FEA-derived screw fixation strength (FS-CL) under craniocaudal cyclic load using machine learning and deep learning, and to explore whether FS-CL can serve as a surrogate marker for pedicle screw loosening (PSL) risk. A retrospective analysis was conducted on 618 screw trajectories data from preoperative of 112 patients. Various ML and DL models utilizing CT images and screw trajectory, were developed to predict screw FS-CL including multilayer perceptron (MLP) and dual-channel 3D ResNet-18 models. Model performance was evaluated using mean squared error (MSE), coefficient of determination (R²) on an external validation set of 126 trajectories. Additionally, we validated the clinical efficiency of the model for the risk assessment of PSL based on a case-control cohort of 62 patients. The MLP and 3D ResNet-18 models demonstrated reliable FS-CL predictions, with less time spent compared to the manual FEA. All DL and ML model that focused on region surrounding screw trajectory performed better. The ResNet-18 model achieved the highest predictive performance for screw FS-CL (MSE: 0.009, R²: 0.836) and highest prediction for PSL risk with an AUC value of 0.826. The MLP model also exhibited moderate performance, outperforming other ML models. AI models proposed in this study can accurately predict FEA-derived FS-CL efficiently providing a supplementary tool for PSL risk evaluation.

OBJECTIVE:Radiologists often face challenges in differentiating benign from malignant sacral bone lesions due to their similar imaging characteristics. This study aimed to develop an ensemble deep learning (DL) model that can preoperatively distinguish between benign and malignant sacral tumors using noncontrast computed tomography images. MATERIALS AND METHODS:Preoperative sacral CT scans from 569 patients with confirmed sacral lesions were analyzed. Data from Center 1 were utilized in model development and internal test via fivefold cross-validation, and those from Centers 2 and 3 were employed in external test. Various ensemble models combining human-readable interpretation and DL were developed. The diagnostic performance of the models and radiologists was assessed using metrics such as precision, recall, accuracy, area under the curve (AUC), F1 score, and confusion matrix. Furthermore, the clinical benefits derived from radiologists' interpretations and supported by the DL model were evaluated. RESULTS:The ensemble model, which integrates 3D-DenseNet121 with human interpretation, exhibited the most robust performance. The ensemble model demonstrated high performance on the internal and external test sets and achieved AUCs of 0.9139 and 0.8713, F1 scores of 0.9054 and 0.8571, precision of 0.9041 and 0.8824, recall of 0.9136 and 0.8333, and accuracy of 0.8630 and 0.8182, respectively. Across the external test cohort, all radiologists experienced improvements in AUC, accuracy, sensitivity, and specificity. Notably, junior radiologists demonstrated significant improvements compared with senior radiologists. CONCLUSION:The potential clinical application of the DL model lies in its capacity to considerably enhance the diagnostic efficiency of radiologists. CRITICAL RELEVANCE STATEMENT:This study presents the first ensemble deep learning model integrating 3D-DenseNet121 with radiologists' interpretation for preoperative differentiation of sacral tumors on noncontrast CT that improved diagnostic performance across all experience levels, particularly for junior radiologists. KEY POINTS:First artificial intelligence-radiologist ensemble for noncontrast computed tomography (NCCT)-based sacral tumor classification. Boosts all radiologists' performance, with the greatest gains for juniors, potentially reducing referrals. Enables reliable NCCT diagnosis, overcoming contrast/magnetic resonance imaging dependency in musculoskeletal oncology.

To evaluate the diagnostic performance of dual-energy computed tomography (DECT) Rho/Z mapping and cinematic rendering (CR) for detecting Achilles tendon rupture. In this prospective study, 117 consecutive patients (median age, 40 years; IQR, 32.8–44.0; range, 17–62) underwent DECT between January and September 2024. MRI served as the reference for rupture diagnosis and quantitative measures (location and gap). For complex/atypical tears (partial tears, avulsion fractures, intratendinous calcifications), operative findings were used as the reference standard. Three independent readers, blinded to the reference, interpreted images in three sessions (grayscale CT, DECT Rho/Z, and CR) to determine rupture presence, location, and gap. Diagnostic metrics (AUC, sensitivity, specificity, accuracy, PPV/NPV) were calculated; Rho/Z cutoffs were derived by ROC analysis. In total, 194 DECT/CR datasets were analyzed (77 bilateral = 154; 40 unilateral = 40). DECT Rho/Z and CR outperformed grayscale CT for rupture detection with AUCs > 0.90 (all P < 0.001). Consensus reader accuracy was higher for DECT (94.9

Background:Knee joint pain is very common in clinical practice, with a complex etiology in which osteoarthritis is the most frequent cause. Among the various types of osteoarthritis, tibiofemoral osteoarthritis (TFOA) is the most prevalent. Patients with lateral patellar compression syndrome (LPCS) also present with knee joint pain. This study aims to compare meniscal and articular cartilage injuries in patients with LPCS and those with TFOA. This study could provide insights into the clinical characteristics and radiological features that distinguish these two conditions. Methods:This study recruited 206 eligible patients from the Department of Sports Medicine at a hospital from March 2018 to February 2023. Patients were divided into two groups of 103: LPCS and TFOA. Magnetic resonance imaging was conducted using standardized protocols. Image analyses were undertaken by experienced radiologists to assess meniscal and cartilage injuries. Results:The mean age was 58.0±10.9 years in the LPCS group and 54.3±10.3 years in the TFOA group, with a significant difference between groups (P=0.01). Patients with LPCS exhibited a higher proportion of meniscal injuries (55.34% vs. 39.81%, P=0.03) and a significantly higher prevalence of posterior root tears of the medial meniscus (34.95% vs. 3.88%, P<0.001) compared with patients with TFOA. Significant differences in the grading of cartilage injuries were observed, particularly in the medial tibiofemoral compartment, where patients in LPCS group had higher grades of injury compared with TFOA patients (P<0.001). Specifically, in the medial compartment, grade 4 cartilage injuries were more frequent in the LPCS group (34.95% vs. 24.27%), while grade 1 injuries were more frequent in the TFOA group (53.40% vs. 29.13%). Among LPCS patients aged ≤50 years, 55% (11/20) had higher-grade cartilage injuries in the lateral compartment than in the medial compartment, suggesting more rapid cartilage damage progression in younger patients with LPCS. Conclusions:This study underscores the significant differences in meniscal and articular cartilage injuries between patients suffering from LPCS and patients with TFOA. The results highlight the importance of radiological features for accurate clinical differentiation and the need for 'tailored' treatment strategies.

Cervical spondylosis is one of the most common degenerative diseases, seriously affecting life quality. Unlike diseases with explicit lesions like cancer, hydroncus, or fracture, the degeneration of the cervical spine cannot be explicitly detected from the appearance of medical images, requiring extensive experience of doctors to interpret subtle clues. However, the extremely high incidence of cervical spondylosis coincides with a serious shortage of experienced doctors and uneven distribution of medical resources, hindering early diagnosis. We propose a cascade-ensemble deep learning framework for cervical spondylosis diagnosis. The framework integrates vertebral body detection and degenerative diagnosis through a cascading architecture, and jointly trains an ensemble of degenerative indicators in a multi-task learning manner. We demonstrate that deep learning models are more sensitive to distance and position based indicators than angle based ones. In intervertebral stenosis analysis, our method achieves comparable performance to senior radiologists and clinicians, with much faster diagnostic speed.

To evaluate intratumoral fat in hepatocellular carcinoma (HCC) using both qualitative and quantitative approaches based on routine chemical-shift magnetic resonance imaging (MRI), and to investigate its potential value in predicting histological grade. This retrospective study included 282 patients with pathologically confirmed HCC between January 2015 and November 2025. Tumors were classified into low-grade and high-grade groups according to the Edmondson–Steiner grade. Intratumoral fat was assessed on in-phase and opposed-phase MRI images. For qualitative assessment, intratumoral fat pattern was categorized as none, heterogeneous, or homogeneous. For quantitative assessment, regions of interest were manually delineated on three consecutive slices showing the largest tumor area, and the mean fat fraction (FF) was calculated. Logistic regression analysis was performed to identify risk factors associated with high-grade HCC. Furthermore, models incorporating clinicoradiological factors were developed for the preoperative prediction of HCC histological grade. Homogeneous intratumoral fat was more frequently observed in low-grade tumors than in high-grade tumors, and FF was significantly higher in low-grade tumors than in high-grade tumors. Both homogeneous intratumoral fat (odds ratio [OR] = 0.230 [0.097–0.514], P = 0.001) and FF (OR = 0.861 [0.811–0.907], P < 0.001) were identified as independent predictors of high-grade HCC. When combined with other clinicoradiological factors, the FF-based model showed better performance than the intratumoral fat pattern–based model (area under the receiver operating characteristic curve: 0.792 vs. 0.744, P = 0.024). Intratumoral fat assessed using chemical-shift imaging provides a simple and noninvasive imaging biomarker for predicting the histological grade of HCC. Both homogeneous intratumoral fat and higher FF were associated with a lower risk of high-grade HCC, and the model based on quantitative assessment outperformed that based on qualitative evaluation.

To develop and validate a clinical-radiomics model based on multiparametric MRI for differentiating solitary primary spinal tumors from solitary spinal metastases. This dual-center retrospective study included 510 patients with pathologically confirmed spinal tumors, randomly split into training (n = 328), internal validation (n = 82), and external test (n = 100) cohorts. Radiomics and deep learning features were extracted from T1-weighted, T2-weighted, T2-weighted fat-suppressed, and T1-weighted contrast-enhanced sequences. Three models were constructed and compared: a radiomics model (Rad-M), a deep learning-radiomics fusion model (DRad-M), and a clinical-radiomics model (CRad-M) that integrated radiomics features with patient age. The performance of seven machine learning classifiers was evaluated for each model. The CRad-M, utilizing a logistic regression (LR) classifier, demonstrated superior performance, achieving areas under the curve (AUCs) of 0.909 and 0.824 on the internal and external test sets, respectively. It significantly outperformed both the Rad-M and DRad-M models (all p < 0.05). The incorporation of deep learning features did not yield a significant improvement over the radiomics-only model. Calibration and decision curve analyses confirmed the robust clinical utility of the CRad-M. The proposed LR-based CRad-M is an effective non-invasive tool for the preoperative differentiation of solitary primary spinal tumors and solitary spinal metastases, with its performance enhanced by the integration of clinical data (age) alongside radiomic features.

OBJECTIVES:To develop a Generative Adversarial Network (GAN) for generating virtual T2 fat-suppressed (T2FS) sequences from standard T1- and T2-weighted images, with the clinical objective of reducing MRI scan time without compromising diagnostic value for spinal tumor assessment. MATERIALS AND METHODS:This retrospective study included 1,389 consecutive patients with spinal tumors from two institutions, divided into training (n = 1,026; 49.2 ± 16.4 years; 540 males), internal validation (n = 257; 48.2 ± 17.2 years; 140 males), and external test (n = 106; 52.8 ± 17.0 years; 59 males) sets. The model used T1- and T2-weighted images as input to generate T2FS images. Quantitative image fidelity evaluations included mean squared error (MSE), structural similarity index measure (SSIM), and peak signal-to-noise ratio (PSNR). The Dice similarity coefficient (DSC) assessed lesion segmentation. Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) measured objective quality. Two experienced radiologists independently rated the images on a 5-point scale, evaluating overall quality, tumor detail preservation, fat suppression performance, and artifacts. RESULTS:The external test set exhibited an MSE of 0.0060 ± 0.0038, SSIM of 0.667 ± 0.120, and PSNR of 23.368 ± 4.298 dB. The real and synthetic images agreed strongly in lesion segmentation, with mean DSC of 0.820 ± 0.176 (internal) and 0.807 ± 0.188 (external). SNR and CNR were comparable between real and synthetic images in both datasets. Qualitative assessments indicated equivalent overall image quality and artifacts. Synthetic images showed superior fat suppression, while real images offered better tumor internal detail. CONCLUSIONS:The proposed GAN-based method generated diagnostically valuable virtual T2FS images. KEY POINTS:1. The proposed deep learning model successfully generated virtual T2FS images from standard T1/T2 MRI, demonstrating favorable quantitative agreement (MSE 0.0060, SSIM 0.667). 2. Synthetic and real images demonstrated strong consistency in lesion segmentation (DSC 0.809-0.824) and comparable SNR/CNR values. 3. While synthetic images provided superior fat suppression, real images maintained slightly better tumor internal detail visualization.

To evaluate the accuracy of deep learning–derived coronary artery calcium scores (DL‑CACS) from non–electrocardiogram (ECG)-gated chest CT against ECG‑gated reference CACS and to determine the impact of CT acquisition parameters on measurement accuracy. From January 2020 to February 2021, 1213 patients at our institution underwent ECG‑gated cardiac CT and non‑gated chest CT within 3 months. An automated pipeline generated DL‑CACS from non‑gated scans. Agreement with ECG‑gated CACS was assessed using Spearman correlation and Bland–Altman analysis; kappa analysis evaluated categorical agreement for the Coronary Artery Calcium Data and Reporting System (CAC‑DRS). Diagnostic performance was evaluated by ROC/AUC. Univariable and multivariable logistic regression quantified associations between acquisition parameters and misclassification. Non‑gated DL‑CACS correlated strongly with gated CACS across LM, LAD, LCX, RCA and TOTAL scores (ρ ≤ 0.865; all p < 0.001). CAC‑DRS agreement was substantial (κ = 0.641), with accuracy 77.5

Accurate preoperative categorization of solitary spinal lesions (SSLs) is important because management differ across lesion categories. This retrospective study included 600 patients with SSLs (332 men, 268 women; median age, 53 years) examined from January 2018 to March 2025 and assigned to training, validation, and test sets using a fixed temporal split. According to histopathology, lesions were categorized as malignant, intermediate, benign, or non-neoplastic. Lesions on CE-T1WI and DWI (b = 50 s/mm²) images from multiple-b-value diffusion-weighted imaging (mb-DWI) were segmented using 3D nnU-Net. Two DenseNet-121 models were developed: a structural MRI model (DL_struct) based on co-registered CE-T1WI, T2WI, and T2-FS images, and a mb-DWI model (DL_mb) based on images acquired at b values of 50, 400, and 800 s/mm² and the corresponding ADC maps. Their late fusion yielded an imaging model (DL_fusion). A clinical model based on five clinico-radiological features was built and integrated with DL_fusion to generate the combined model. Performance was compared with five radiologists. In the test cohort, Dice scores were 0.900 for CE-T1WI and 0.857 for DWI (b = 50 s/mm²). The combined model achieved the best four-category classification performance, with a macro-AUC/accuracy of 0.921/0.772, compared with 0.916/0.743 for DL_fusion, 0.852/0.640 for DL_struct, 0.871/0.647 for DL_mb, and 0.721/0.493 for the clinical model. Category-specific AUCs ranged from 0.893 to 0.956. The combined model achieved an accuracy of 0.772, outperforming the junior and attending radiologists (0.588–0.750), although it remained inferior to the senior radiologist (0.853). Model assistance improved reader performance, with the largest gains in junior radiologists. This multimodal DL framework enabled automated segmentation and four-category classification of SSLs and improved the performance of less experienced readers.

To investigate the performance of a free-breathing coronary computed tomography angiography (CCTA) protocol using a high-threshold bolus-tracking strategy in improving coronary enhancement and diagnostic visualization in patients with delayed contrast arrival. In this prospective study, 159 patients undergoing CCTA were randomized to either a conventional bolus-tracking protocol (breath-hold, 100 HU trigger threshold, 6.8 s delay) or a modified high-threshold protocol (free-breathing, 300 HU trigger threshold, 2 s delay). Patients were stratified according to contrast arrival time (Tarr) into normal and delayed contrast arrival groups using a predefined operational cutoff (15 s), resulting in four subgroups. Coronary attenuation, enhancement uniformity as assessed by the coefficient of variation (COV), enhancement phase distribution, and image quality scores were compared among groups. Under the conventional protocol, patients with delayed contrast arrival demonstrated reduced enhancement uniformity and a higher proportion of late-phase acquisitions. Compared with the conventional protocol, the modified protocol significantly increased coronary attenuation and improved enhancement uniformity (COV: 11.28

Background: Risk stratification of spinal tumors is a major unmet clinical need for personalized therapy. Purpose: To explore the feasibility of pretreatment whole-lesion apparent diffusion coefficient (ADC) histogram in predicting local recurrence of aggressive spinal tumors. Methods: 119 aggressive spinal tumor patients (median age, 40; range, 13-74 years) confirmed by pathological findings with a mean follow-up of 36 months were enrolled and divided into the recurrence and non-recurrence group. The histogram metrics of whole-lesion, including the maximum, mean, kurtosis, skewness, entropy, and percentiles (10th, 25th, 50th, 75th, 95th) ADC values, were evaluated and take the average. Fractal dimension (FD) was assessed in the three orthogonal directions and take maximum. Clinical and general imaging features were used to construct an alternative prognostic model for comparison. Variables with statistical differences would be included in stepwise logistic regression analysis. Results: As for the clinical model, Enneking staging (odds ratio [OR]: 3.572; P = 0.04) and vertebral compression (OR: 4.302; P = 0.002) were independent predictors of recurrence. There was no statistical difference in FD between the two groups (P = 0.623). Among the ADC histogram parameters compared, skewness, maximum, and mean ADC values were independent risk factors and constructed ADC histogram prediction models. The ADC histogram model (AUC = 0.871) and the combined model (AUC = 0.884) performed better than the clinical prediction model (AUC = 0.704) with P-values of 0.004 and 0.001, respectively. Conclusion: Prediction models based on the ADC histogram analysis might represent serviceable instruments for the aggressive spinal tumors.

Background:Metal artifacts (MAs) induced by dental prostheses in carotid computed tomography angiography (CTA) significantly impair diagnostic accuracy. This study aimed to assess the efficacy of the iterative metal artifact reduction (iMAR) technique in mitigating these artifacts. Methods:Eighty-one patients with suspected vascular disorders and dental prostheses who underwent CTA imaging were retrospectively included. The CTA images were reconstructed with and without iMAR (iMAR-CTA and non-iMAR-CTA) for evaluation. Additionally, 81 matched patients without dental prostheses who underwent CTA imaging (standard CTA) served as a reference group for objective image quality assessment. Objective image quality involving signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and artifact index (AI) were analyzed. Subjective image quality was evaluated using a five-point Likert scale. Diagnostic performance was assessed by examining luminal stenosis, calcification, and aneurysm, with digital subtraction angiography (DSA) as the reference standard. Intramodality and inter-radiologist agreements were calculated using the intraclass correlation coefficient (ICC). Results:Image quality score was significantly higher for iMAR-CTA images than non-iMAR-CTA images [radiologist 1, 5 (5-5) vs. 3 (2-3); radiologist 2, 5 (4-5) vs. 3 (3-3); radiologist 3, 5 (5-5) vs. 2 (2-3), all P<0.001]. There was no significant difference in scores between iMAR-CTA and normal CTA. In the objective analysis, iMAR-CTA exhibited higher SNR and CNR and lower AI compared to non-iMAR-CTA (P<0.001). Furthermore, the objective image quality of iMAR-CTA was comparable to that of standard CTA, with no statistically significant differences in SNR (P=0.324) or CNR (P=0.109). For diagnostic performance evaluation, iMAR-CTA exhibited good to excellent agreement with DSA for luminal stenosis and aneurysm (ICC, 0.859-0.946), exceeding the moderate to good agreement of non-iMAR-CTA (ICC, 0.583-0.777). Regarding luminal stenosis severity, iMAR-CTA had higher accuracy rates (90.63-93.75%; 58/64-60/64) than non-iMAR-CTA (57.81-65.63%; 37/64-42/64). In aneurysm detection, iMAR-CTA achieved higher accuracy rates (77.78-88.89%; 7/9-8/9) than non-iMAR-CTA (44.44-66.67%; 4/9-6/9). For luminal stenosis severity and calcification, iMAR-CTA demonstrated excellent agreement (ICC, 0.908-0.910), whereas non-iMAR-CTA exhibited moderate agreement (ICC, 0.694-0.747). Conclusions:iMAR effectively reduces MAs, achieving image quality comparable to standard CTA without artifacts, facilitating a more reliable evaluation of carotid artery disorders in patients with dental prostheses.

BACKGROUND:Sarcopenia is a prevalent comorbidity in patients with chronic obstructive pulmonary disease (COPD). We aimed to investigate the impact of sarcopenia diagnosed by chest CT on mortality in critically ill patients with exacerbation of COPD (ECOPD). METHODS:This retrospective study enrolled 148 patients hospitalized in the intensive care unit due to ECOPD from 2018 to 2023. Sarcopenia was defined by the skeletal muscle index measured at the 12th thoracic vertebra (T12) level on chest CT. Patients were categorized into the sarcopenia and non-sarcopenia groups. Hospitalization duration, short-term (30 and 90-day) and long-term (1-year and overall) COPD-related mortality and all-cause mortality were compared between the two groups. Cox regression analyses were conducted to recognize the risk factors for mortality, and a sarcopenia-based nomogram was developed. RESULTS:Eighty-four patients (56.76 %) with sarcopenia were identified through chest CT measurements. The 1-year COPD-related and all-cause mortality, as well as overall COPD-related and all-cause mortality, were significantly higher in the sarcopenia group than the non-sarcopenia group (19.05 % vs. 4.69 %, p = 0.010; 28.57 % vs. 6.25 %, p = 0.001; 33.33 % vs. 15.63 %, p = 0.015; 47.62 % vs. 29.69 %, p = 0.027, respectively). Multivariate Cox regression analyses revealed sarcopenia as a risk factor for 1-year (HR = 3.981 [1.137-13.938], p = 0.031) and overall (HR = 2.308 [1.310-4.065], p = 0.004) mortality. The sarcopenia-based nomogram demonstrated favorable prognostic performance. CONCLUSIONS:Sarcopenia evaluated at the T12 level on chest CT may serve as a prognostic factor for predicting long-term mortality among critically ill patients with ECOPD.

Retroperitoneal sarcoma (RPS) is highly heterogeneous, leading to different risks of distant metastasis (DM) among patients with the same clinical stage. This study aims to develop a quantitative method for assessing intratumoral heterogeneity (ITH) using preoperative contrast-enhanced CT (CECT) scans and evaluate its ability to predict DM risk. We conducted a retrospective analysis of 274 PRS patients who underwent complete surgical resection and were monitored for ≥ 36 months at two centers. Conventional radiomics (C-radiomics), ITH radiomics, and deep-learning (DL) features were extracted from the preoperative CECT scans and developed single-modality models. Clinical indicators and high-throughput CECT features were integrated to develop a combined model for predicting DM. The performance of the models was evaluated by measuring the receiver operating characteristic curve and Harrell’s concordance index (C-index). Distant metastasis-free survival (DMFS) was also predicted to further assess survival benefits. The ITH model demonstrated satisfactory predictive capability for DM in internal and external validation cohorts (AUC: 0.735, 0.765; C-index: 0.691, 0.729). The combined model that combined clinicoradiological variables, ITH-score, and DL-score achieved the best predictive performance in internal and external validation cohorts (AUC: 0.864, 0.801; C-index: 0.770, 0.752), successfully stratified patients into high- and low-risk groups for DM (p < 0.05). The combined model demonstrated promising potential for accurately predicting the DM risk and stratifying the DMFS risk in RPS patients undergoing complete surgical resection, providing a valuable tool for guiding treatment decisions and follow-up strategies. The intratumoral heterogeneity analysis facilitates the identification of high-risk retroperitoneal sarcoma patients prone to distant metastasis and poor prognoses, enabling the selection of candidates for more aggressive surgical and post-surgical interventions.

Recent advances in generative latent space sampling for enhanced generation quality have demonstrated the benefits from the Energy-Based Model (EBM), which is often defined by both the generator and the discriminator of off-the-shelf Generative Adversarial Networks (GANs) of many types. However, such latent space sampling may still suffer from mode dropping even sampling in a low-dimensional latent space, due to the inherent complexity of the data distributions with rugged energy landscapes. Motivated by the success of Wang-Landau (WL) sampling in statistical physics, we propose WL-GAN, a collaborative learning framework for generative latent space sampling, where both the invariant distribution and the proposal distribution of the Markov chain are jointly learned on the fly, by exploiting the historical statistics behind the simulated samples. We show that the two learning modules work together for better balance between exploration and exploitation over the energy space in GAN sampling, alleviating mode dropping and improving the sample quality of GAN. Empirically, the efficacy of WL-GAN is demonstrated on both synthetic datasets and real-world image datasets, using multiple GANs. Code is available at https://github.com/zeyihou/collaborative-learn.

We aimed to develop and internally validate a radiomics classification model based on multiphase computed tomography (CT) scans for preoperative differentiation of retroperitoneal non-fatty dedifferentiated liposarcoma (DDL) from leiomyosarcoma (LMS). This retrospective study enrolled 78 DDL patients and 51 LMS patients who underwent surgical resection and pathological confirmation at our hospital between January 2011 and April 2023. Enhanced CT scans were performed within two weeks prior to surgery. An experienced radiologist manually delineated the tumor regions of interest using ITK-SNAP software on arterial-phase CT images, with contours copied to plain and venous-phase images. The dataset was split hierarchically (80

Objective. Ossification of the posterior longitudinal ligament (OPLL) is a prevalent cervical spine degeneration disease leading to significant spinal cord dysfunctions. Due to morphological diversity and data scarcity, traditional OPLL assessment relies on manual measurements, which suffer from low consistency and high cost. To implement automated quantification of the OPLL, a cognition-inspired segmentation framework, named the probabilistic anatomical cognition (PAC) framework, is proposed to encode physicians' anatomical knowledge of the OPLL and mimic their hierarchical logic of inferring lesions. Approach. The OPLL anatomical structure is firstly modeled by a multi-level probabilistic representation from the stochastic global shape of the spinal canal (SC) to the local feature distributions of the lesions. Based on the anatomical prior model, the OPLL segmentation is implemented by the deep-logic shape inference. The logic extracts high-confidence global feature observations of the SC, following with the inference to the local lesions by morphological correlations. The fusion of the anatomical prior and multi-level observations enhances both interpretability and generalization of lesion segmentation and reduces reliance on large datasets. Main results. Tested on a clinical dataset of 439 patients, the PAC framework improves dice similarity coefficient by 10% over the lightweight baseline and achieves high consistency with expert assessments on clinical lesion metrics. Significance. A general automated segmentation pipeline and three-dimensional metrics are provided for the first time by the framework to quantify the OPLL degeneration, which offers valuable insights to support surgical decision-making.

BACKGROUND:Endometrial cancer (EC) is a common gynecologic malignancy; accurate assessment of key prognostic factors is important for treatment planning. PURPOSE:To develop a deep learning (DL) framework based on biparametric MRI for automated segmentation and multitask classification of EC key prognostic factors, including grade, stage, histological subtype, lymphovascular space invasion (LVSI), and deep myometrial invasion (DMI). STUDY TYPE:Retrospective. SUBJECTS:A total of 325 patients with histologically confirmed EC were included: 211 training, 54 validation, and 60 test cases. FIELD STRENGTH/SEQUENCE:T2-weighted imaging (T2WI, FSE/TSE) and diffusion-weighted imaging (DWI, SS-EPI) sequences at 1.5 and 3 T. ASSESSMENT:The DL model comprised tumor segmentation and multitask classification. Manual delineation on T2WI and DWI acted as the reference standard for segmentation. Separate models were trained using T2WI alone, DWI alone and combined T2WI + DWI to classify dichotomized key prognostic factors. Performance was assessed in validation and test cohorts. For DMI, the combined model's was compared with visual assessment by four radiologists (with 1, 4, 7, and 20 years' experience), each of whom independently reviewed all cases. STATISTICAL TESTS:Segmentation was evaluated using the dice similarity coefficient (DSC), Jaccard similarity coefficient (JSC), Hausdorff distance (HD95), and average surface distance (ASD). Classification performance was assessed using area under the receiver operating characteristic curve (AUC). Model AUCs were compared using DeLong's test. p < 0.05 was considered significant. RESULTS:In the test cohort, DSCs were 0.80 (T2WI) and 0.78 (DWI) and JSCs were 0.69 for both. HD95 and ASD were 7.02/1.71 mm (T2WI) versus 10.58/2.13 mm (DWI). The classification framework achieved AUCs of 0.78-0.94 (validation) and 0.74-0.94 (test). For DMI, the combined model performed comparably to radiologists (p = 0.07-0.84). CONCLUSIONS:The unified DL framework demonstrates strong EC segmentation and classification performance, with high accuracy across multiple tasks. EVIDENCE LEVEL:3. TECHNICAL EFFICACY:Stage 3.

Spinal metastasis surgery frequently results in anemia, affecting patient recovery, yet lacks a quantitative method for assessing the risk of postoperative anemia. This study investigates the potential of MRI-based radiomics models to predict postoperative anemia, aiding in personalized treatment and improved outcomes. 247 patients diagnosed with spinal metastases pathologically and underwent surgery from December 2012 to December 2023 were enrolled and divided into postoperative anemia (n = 158) and non-anemia (n = 89) groups. Radiomics features were extracted from regions of interest on sagittal T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), and fat-suppressed (FS)-T2WI sequences of preoperative MRI scans. Then, seven radiomics models were developed using logistic regression analysis, supported by a five-fold cross-validation technique. The Radscore, originating from the model with the highest predictive accuracy, was chosen for nomogram development. After variable selection via stepwise logistic regression analyses, clinical variables and the Radscore were included in the clinical and combined clinical and Radscore models. Ultimately, three models—clinical, Radscore, and combined clinical and Radscore models—were developed. Receiver operating characteristic analyses, Brier score, calibration curves, and decision curve analyses were used for model performance evaluation. Among the radiomics models, the one with feature integration based on T1WI and FS-T2WI sequences performed the best, with area under the curve (AUC) values of 0.844 (95