Background:Breast cancer is a heterogeneous disease, and accurate subtype characterization is essential for guiding personalized treatment. In particular, HER2-low tumors have recently emerged as a distinct clinical entity with potential responsiveness to novel HER2-targeted therapies. However, reliable noninvasive imaging methods to identify these subgroups remain lacking. Purpose:To evaluate the potential of time-dependent diffusion MRI (Td-dMRI) in differentiating breast cancer molecular subtypes and to investigate its correlation with immunohistochemical biomarkers, particularly the newly established three-tier HER2 classification. Materials and methods:In this retrospective study, female patients with untreated invasive ductal carcinoma underwent 3T breast MRI including Td-dMRI between June 2023 and October 2024. A custom protocol combining oscillating gradient spin-echo (OGSE) and pulsed gradient spin-echo (PGSE) sequences enabled diffusion sampling at multiple diffusion times and frequencies. Microstructural parameters-cellularity, extracellular and intracellular diffusivity (Dex, Din), cell diameter, intracellular volume fraction (fin), and intracellular water residence time (τin)-were estimated using a Bayesian model based on a joint multicompartmental framework. Molecular subtypes (Luminal A/B, HER2-enriched, triple-negative [TN]) and HER2 expression levels (HER2-zero, HER2-low, HER2-positive) were determined via IHC and fluorescence in situ hybridization (FISH). Quantitative Td-dMRI metrics were compared across subtypes and correlated with ER, PR, HER2, and Ki-67 status using ANOVA, Kruskal-Wallis, and ROC curve analysis. Results:This study included 71 female participants (mean age, 51.3 ± 10.2 years). Multiple Td-dMRI parameters varied significantly across molecular and HER2 subtypes. ADC50Hz was significantly higher in Luminal A compared to Luminal B (P = 0.003). HER2-enriched tumors showed higher ADC values and cell diameters but lower cellularity compared to Luminal B (P< 0.05). ER- and PR- tumors had higher ADCs, cell diameters, and Din, with lower cellularity than positive counterparts. Din effectively distinguished TN from non-TN cancers (AUC = 0.710). For HER2 stratification, ADC30ms distinguished HER2-zero from HER2-low tumors with high accuracy (AUC = 0.898), and cell diameter and cellularity were most effective for differentiating HER2-low from HER2-positive tumors (AUC = 0.770). No significant Td-dMRI differences were observed for Ki-67. Conclusion:ADC30ms most effectively distinguished HER2-zero from HER2-low tumors, while microstructural parameters such as cellularity and cell diameter moderately differentiated HER2-low from HER2-positive cancers. These results support the potential of Td-dMRI as a complementary imaging biomarker for subtype characterization, although findings were limited by small subgroup sizes and the single-center design.

To investigate the feasibility of 3D contrast-enhanced CT radiomics features to predict response to neoadjuvant chemotherapy (NAC) for adenocarcinoma of the esophagogastric junction (AEG) and to develop and validate a nomogram to assist in clinical decision-making. The clinical, pathological, and CT data of 239 patients with locally advanced AEG who underwent NAC and radical resection were retrospectively collected between March 2016 and June 2023 from two independent Chinese medical centers. They were randomly assigned to a training cohort, an internal verification cohort, or an external verification cohort. Based on the CT radiomics features after dimension reduction, the radiomics model was constructed using linear discriminant analysis as the classifier to obtain the radiomics score. Clinical characteristics were screened, and multivariable logistic regression was applied to construct the clinical model. The combined model was generated by integrating clinical features and radiomics scores, upon which a nomogram was subsequently developed. Finally, receiver operating characteristic curves, calibration curves, and decision curves were plotted to evaluate the predictive performance, calibration performance, and clinical benefits of each model for the efficacy of NAC in AEG patients. Overall, 86 of the 239 patients responded well to NAC. The nomogram was comprised of tumor thickness, lymph node short diameter, and the radiomics score. In the training cohort, the AUC values of the clinical model, the radiomics model, and the combined model for predicting NAC response were 0.771 (95

Background: Accurate noninvasive prediction of pathological complete response (pCR) after neoadjuvant chemotherapy (NAC) in invasive breast cancer (BC) remains challenging. This study aimed to develop and validate a multivariable prediction model integrating clinicopathological variables, immunoinflammatory markers, and multiparametric magnetic resonance imaging (MRI) features for predicting pCR after NAC. Methods: In this retrospective multicenter study, 345 women with invasive BC who underwent pretreatment breast MRI and NAC were included. pCR was defined as the absence of residual invasive cancer in the breast and axillary lymph nodes at surgery. Patients were divided into training and internal validation cohorts, with an independent external cohort used for validation. Clinicopathological variables, immunoinflammatory markers, and MRI features, including mean apparent diffusion coefficient (ADCmean), were collected. Predictor selection was performed using the least absolute shrinkage and selection operator and multivariable logistic regression. Model performance was assessed using receiver operating characteristic analysis. Results: Among the 345 patients, 116 (33.62%) achieved pCR. Clinical T stage, lymphocyte-to-monocyte ratio (LMR), systemic immune-inflammation index (SII), enhancement pattern, and ADCmean were independent predictors. The combined model showed the best performance, with area under the curves of 0.820, 0.810, and 0.799 in the training, internal validation, and external validation cohorts, outperforming the clinical and MRI models. Conclusions: The combined model integrating clinicopathological variables, immunoinflammatory markers, and multiparametric MRI features may help predict pCR after NAC and support individualized treatment planning and potential surgical de-escalation.

[This corrects the article DOI: 10.3389/fonc.2025.1739008.].

OBJECTIVES:This study aimed to compare image quality and diagnostic performance between artificial intelligence-assisted compressed sensing (ACS) images reconstructed using deep learning reconstruction (ACS-DLR) and conventional parallel imaging (PI) images in rectal cancer MRI. METHODS:107 patients with biopsy-proven rectal cancer were included. MRI included conventional PI and ACS acquisitions, with the ACS raw data reconstructed at three deep learning reconstruction strength levels (ACS-L, ACS-M, and ACS-H). Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were compared across four image sets using the Friedman test. Subjective image quality was assessed using a 5-point Likert scale for overall image quality, noise, artefact, and edge sharpness. Interobserver agreement for objective metrics was measured by ICC, and for subjective metrics by Cohen's kappa. Diagnostic performance was evaluated using postoperative histopathology, including T stage, N stage, extramural venous invasion (EMVI), and mesorectal fascia (MRF) involvement. RESULTS:ACS reduced acquisition time by 50 % (from 3 min 20 s to 1 min 40 s). Lesion SNR did not differ significantly among the four image sets (P > 0.05), but ACS-H showed the highest muscle SNR. CNR showed significant differences in selected pairwise comparisons. ACS-H achieved the highest subjective scores for overall image quality, noise reduction, and lesion edge sharpness. In the surgical subcohort, ACS-H improved T staging accuracy (P = 0.010; P = 0.018), MRF involvement assessment (P = 0.004; P = 0.012), and EMVI sensitivity (P = 0.039; P = 0.041). N staging accuracy was not significantly different (P = 0.521; P = 0.841). CONCLUSION:ACS reduced acquisition time, while ACS-DLR improved subjective image quality. ACS-H improved T-stage and MRF assessment and increased EMVI sensitivity, whereas N-stage accuracy did not improve significantly.

PURPOSE:To evaluate the diagnostic performance of the Liver Imaging Reporting and Data System (LI-RADS) nonradiation Treatment Response Assessment (TRA) version 2024 (v2024) in assessing treatment response of hepatocellular carcinoma (HCC) following downstaging locoregional therapy (LRT) plus immune-targeted therapy; to compare its diagnostic performance with LI-RADS TRA version 2018 (v2018); and to explore the added diagnostic value of its ancillary features (AFs). METHODS:This retrospective study included patients with HCC who underwent LRT plus immune-targeted therapy followed by hepatectomy between January 2021 and December 2025. Two radiologists independently assigned LR-TR classifications to lesions on MRI according to the LI-RADS Nonradiation TRA v2024 and v2018. With pathological findings from hepatectomy as the reference standard, the sensitivity, specificity and accuracy of the two versions were calculated. Diagnostic performance for identifying residual tumor viability was compared using the McNemar test. RESULTS:A total of 132 patients (19 women, 113 men; mean age, 55.9 ± 9.5 years [standard deviation]) with 143 treated HCC lesions were enrolled in this study. The LR-TR Viable category of v2024 yielded a sensitivity of 66.3% (95% confidence interval [CI]: 57.1%-75.5%) and a specificity of 95.2% (95% CI: 84.2%-98.7%) for predicting pathologic viability. For identifying complete pathologic necrosis, the LR-TR Nonviable category showed a sensitivity of 76.2% (95% CI: 63.3%-89.1%) and a specificity of 95.1% (95% CI: 90.8%-99.3%). The incorporation of AFs significantly improved the sensitivity for detecting pathologic viability (87.1% vs. 66.3%, P < 0.001), with no significant difference in specificity (90.5% vs. 95.2%, P = 0.500). Overall, LI-RADS nonradiation TRA v2024 integrated with AFs exhibited superior diagnostic performance relative to LI-RADS TRA v2018. CONCLUSION:The LI-RADS nonradiation TRA v2024 incorporating AFs showed excellent diagnostic performance and outperformed LI-RADS TRA v2018 for assessing tumor viability in HCC patients treated with LRT plus immune-targeted therapy.

Background Tertiary lymphoid structures (TLS) are spatially organized immune niches associated with therapeutic response and favorable outcomes in breast cancer (BC). However, TLS assessment currently relies on invasive tissue-based analyses, and the biological mechanisms underlying imaging-based TLS prediction remain poorly understood.Methods We developed and validated a spatial heterogeneity-based radiomic TLS signature (shTLS) using dynamic contrast-enhanced MRI to non-invasively predict TLS status across multicenter BC cohorts. Spatial habitat radiomics were used to capture intratumoral and peritumoral immune-related heterogeneity. Integrated multi-omics analyses, including transcriptomics, pathomics, genomics, single-cell RNA sequencing, immunohistochemistry, and multiplex immunofluorescence, were performed to biologically interpret shTLS-defined subgroups. Functional drug-sensitivity assays were conducted to assess therapeutic implications.Results The shTLS model achieved robust predictive performance across independent cohorts and molecular subtypes. High shTLS scores were associated with immune-inflamed tumors characterized by spatially clustered activated T cells and dendritic cells (DCs). In contrast, shTLS-low tumors exhibited an immunosuppressive spatial niche with peripheral accumulation of CD4+ PD-1+ T cells and plasma cells, increased immune-tumor separation, and enhanced inflammatory and immunoregulatory signaling. An indoleamine 2,3-dioxygenase 1 (IDO1)-associated immunoregulatory program was observed in the shTLS-low tumors, which appeared to be preferentially expressed by LAMP3+CCR7+ migratory DCs. Pharmacologic inhibition of IDO1 enhanced chemotherapy and CDK4/6 inhibitor sensitivity in vitro.Conclusion This study establishes spatial radiomics as a non-invasive approach to decode TLS-associated immune ecosystems and supports the presence of an IDO1-associated immunosuppressive phenotype, providing biological insight and translational rationale for patient stratification and future combination strategies.

Early prediction of pathological complete response (pCR) to neoadjuvant chemotherapy (NAC) in breast cancer remains challenging. Traditional imaging lacks molecular sensitivity, while single-modality radiomics may miss tumor microenvironment dynamics. Integrating amide proton transfer-weighted imaging (APTWI)—quantifying protein levels linked to chemo-response—into multimodal radiomics may enable precise early pCR prediction, guiding personalized therapy. This study aims to develop and validate a multimodal radiomics model, integrating amide proton transfer-weighted imaging (APTWI), diffusion-weighted imaging (DWI), and early-phase contrast-enhanced T1WI, for preoperative prediction of pathological complete response (pCR) in breast cancer patients undergoing neoadjuvant chemotherapy (NAC). This retrospective study included 109 women (mean age 50 ± 10 years) with untreated breast cancer between May 2023 and August 2024, underwent NAC and pretreatment MRI scanning, including APTWI, DWI, and dynamic contrast-enhanced T1WI. Three-dimensional tumor segmentation was performed using ITK-SNAP. Radiomics features were extracted from APT, ADC maps, and enhanced subtraction images (30s/90s) using PyRadiomics, adhering to IBSI guidelines. LASSO regression selected predictive features, followed by Support Vector Machine and Logistic Regression classifiers with five-fold cross-validation. Performance was evaluated via ROC analysis, calibration curves, decision curve analysis (DCA), and SHAP interpretability. Patients were randomly divided into training (n = 77) and testing (n = 32) cohorts. The integrated model combining clinical (e.g. HER2, Ki-67) and radiomics features achieved superior predictive performance (AUC = 0.888, 95

Background:Undifferentiated pleomorphic sarcoma (UPS) is an aggressive soft tissue malignancy in which the accurate assessment of histological grade is crucial for treatment planning and prognosis. Noninvasive magnetic resonance imaging (MRI)-based tumor features may reflect tumor biology, but their association with histological grade and survival outcomes in UPS remains unclear. The aim of this study was to investigate the relationship between MRI-derived tumor features, histological grade, and survival outcomes in patients with UPS, in order to evaluate the potential of MRI as a noninvasive tool for prognostication and treatment planning. Methods:This retrospective study included 83 patients with pathologically confirmed UPS between January 2015 and December 2023. All patients underwent pre-treatment 3.0 Tesla (3T) MRI scans, which included T1-weighted, T2-weighted, and diffusion-weighted imaging (DWI). The MRI features assessed included growth pattern, signal intensity (SI) heterogeneity, necrosis volume, and apparent diffusion coefficient (ADC) values. Tumor histological grade was determined using the Fédération Nationale des Centres de Lutte Contre le Cancer (FNCLCC) system. Statistical analyses, including univariate and multivariate logistic regression, were performed to identify imaging features associated with high-grade UPS. Results:The study included 83 patients with a mean age of 59.5 years, and 67.5% had high-grade (Grade II-III) tumors. Significant MRI features associated with high-grade UPS included diffuse infiltrative growth pattern (P=0.047), high T2-weighted SI heterogeneity (P=0.04), tumor volume with necrosis ≥50% (P=0.03), and lower ADC mean values (P=0.001). Multivariable analysis revealed that growth pattern, necrosis volume, and ADC difference were independent predictors of high-grade tumors. The combination of these features had a high diagnostic accuracy, with an area under the curve (AUC) of 0.876, sensitivity of 82.14%, and specificity of 85.19%. Conclusions:MRI features, including growth pattern, necrosis-related signal, and ADC values, are significantly associated with histological grade in UPS.

To investigate the predictive value of habitat imaging based on spectral CT-derived iodine maps (IMs) for pathologic response to neoadjuvant therapy (NAT) in gastric cancer (GC). This retrospective, two-center study included 151 patients with pathologically confirmed GC who underwent NAT followed by gastrectomy between July 2022 and June 2025. All patients underwent dual-phase, contrast-enhanced, dual-layer spectral CT scans before NAT. Based on tumor regression grade, patients were categorized as responders or non-responders. In venous-phase IMs, tumor voxels in the entire lesion were clustered into distinct habitats using the k-means algorithm. The volume fraction of each habitat and an intratumoral heterogeneity score (ITHscore) were calculated. Univariate analysis and logistic regression analyses determined predictive parameters among clinicopathologic and IM-based variables. A weighted logistic regression model for responders was developed and validated using fivefold cross-validation and an external test set. Among the 151 patients, 54 (35.8

BACKGROUND:Preoperative immuno-chemotherapy improves outcomes in triple-negative breast cancer, but associated imaging response patterns remain poorly characterized. This study aimed to describe a novel MRI phenomenon, tumor flare-like response (TFLR), and evaluate its association with pathological complete response (pCR). METHODS:Ad hoc imaging analysis of a prospective phase II trial (NCT04213898; n = 39). Breast MRI was performed at baseline and after every two cycles. TFLR was defined a priori as new enhancing nodules (≥5 mm) separate from the index tumor on DCE-MRI, appearing during neoadjuvant therapy with subsequent regression. Inter-reader agreement was assessed with Cohen's kappa and intraclass correlation coefficient. Univariate logistic regression and ROC analysis (Youden index) with 1000-bootstrap resampling identified the optimal largest-nodule size cut-off for pCR prediction. Two multivariable logistic regression models (continuous-size and binary-size) evaluated independence after adjustment for PD-L1 status, FGT category, baseline tumor size, and clinical stage; multicollinearity was assessed by variance inflation factor. RESULTS:TFLR occurred in 74.4% (29/39) of patients and was significantly associated with higher FGT density (P = .03). Among the 29 patients with TFLR, nodules were predominantly oval (93.1%) with circumscribed margins, homogeneous enhancement. Distribution was bilateral (72.5%), ipsilateral (17.2%), or contralateral (10.3%) relative to the primary tumor, with consistently asymmetric counts between breasts. Kinetic curves were persistent (31.0%), plateau (27.6%), or washout (41.4%). TFLR first appeared after cycle 2 in all cases and completely resolved in 75.9% (22/29) by treatment completion; median persistence was 152 days (95% CI 145-159). Presence of TFLR alone did not predict pCR (P = .72). Patients with largest nodule ≥9 mm had significantly higher pCR rate (77.8% vs. 36.4%, P = .048). ROC analysis identified 9 mm as the optimal cut-off. In multivariable analysis, largest nodule diameter remained independently predictive whether analyzed continuously (OR = 2.965, 95% CI 1.298-6.772, P = .01) or dichotomized at ≥9 mm (OR = 7.833, 95% CI 1.260-48.701, P = .027). Model AUCs were 0.904 (95% CI 0.797-1.00) and 0.727 (95% CI 0.519-0.936), respectively. CONCLUSIONS:TFLR is a frequent, reversible MRI finding in TNBC treated with camrelizumab-based neoadjuvant immuno-chemotherapy. Largest nodule diameter ≥9 mm is a strong, independent predictor of pathological complete response and represents a promising early, non-invasive imaging biomarker of immunotherapeutic efficacy.

BACKGROUND:Morphological changes of abdominal organs in hepatocellular carcinoma (HCC) remain uncharacterized. This study aimed to automatically quantify these changes using deep learning, assess their treatment-related outcome associations, and evaluate prognostic value. METHODS:Abdominal computed tomography (CT) images from 2,747 patients with HCC and 2,869 healthy controls across 15 cohorts from 10 centers (8 Chinese and 2 international) were analyzed. The patients included those receiving immune checkpoint inhibitors (ICIs), transarterial chemoembolization (TACE), or surgical resection. A deep learning algorithm automatically segmented the spleen, liver, kidneys, pancreas, and adrenal glands. Organ volumes and height-normalized indexes (volume/height2) were calculated. Propensity score matching balanced baseline differences. Cox regression assessed associations with overall survival (OS), progression-free survival (PFS), and disease-free survival (DFS), with subgroup and interaction analyses. FINDINGS:Compared with healthy controls, patients with HCC showed significant organ remodeling, with enlarged spleen, liver, kidneys, and adrenal glands (all p < 0.001). Multivariable analysis showed that in the ICI cohort, left kidney volume/index, right kidney index, and left adrenal gland volume/index predicted longer OS and PFS; in the TACE cohort, spleen volume/index and left kidney volume predicted OS and PFS; and in the surgical cohort, spleen volume, left kidney volume, left adrenal gland volume/index, and liver index were independent predictors. Restricted cubic spline analysis suggested nonlinear relationships between adrenal and kidney volumes and survival. CONCLUSIONS:HCC is associated with systemic abdominal organ remodeling. Automated CT-based multi-organ quantification offers reproducible, non-invasive prognostic biomarkers, particularly in cases involving adrenal glands, spleen, and kidneys, supporting personalized treatment and prognosis assessment. FUNDING:This work was funded by the National Natural Science Foundation of China.

Background:The hepatobiliary phase (HBP) of gadoxetic acid-enhanced liver magnetic resonance imaging (MRI) is important for detecting colorectal liver metastasis (CRLM), but image quality may be limited. This study evaluated whether deep learning-based reconstruction united compressed sensing (DR-uCS) and deep learning-based reconstruction high-resolution united compressed sensing (DR-HR-uCS) improve image quality and lesion detection in CRLM. Methods:This retrospective study included 86 patients with 116 CRLM lesions (71 lesions ≥1 cm and 45 lesions <1 cm) who underwent 3.0-T gadoxetic acid-enhanced liver MRI. A standard-resolution HBP acquisition was reconstructed into conventional united compressed sensing (uCS) and DR-uCS from the same raw k-space data, while a separate high-resolution acquisition generated DR-HR-uCS images. Two radiologists independently assessed subjective image quality, artifact severity, liver edge/vessel clarity, and lesion conspicuity. Quantitative metrics [liver signal-to-noise ratio (SNR), lesion SNR, and contrast-to-noise ratio (CNR)] were measured by standardized region-of-interest analysis. Diagnostic performance for lesions ≥1 and <1 cm was evaluated using pathology or multidisciplinary consensus. Diagnostic time was recorded across three reader experience levels. Results:Both DR-uCS and DR-HR-uCS significantly improved overall image quality compared with uCS (median score: 5 vs. 4, both P<0.001) and significantly reduced image artifacts (both P<0.001). DR-uCS achieved the highest liver SNR and CNR, while the lesion SNR was comparable across methods (P=0.03 and P=0.001, respectively). For lesions ≥1 cm, conspicuity and diagnostic performance were similar (all P>0.05). For lesions <1 cm, DR-HR-uCS demonstrated higher conspicuity and sensitivity (87.1%) than uCS (72.4%) and DR-uCS (78.6%) (adjusted P<0.05), with comparable specificity. Diagnostic time for sub-centimeter lesions was significantly shorter with DR-HR-uCS (P<0.001), and differences among readers were reduced. Conclusions:DR-uCS improves HBP image quality, while DR-HR-uCS further enhances the detection efficiency and conspicuity of sub-centimeter CRLMs. Its advantage likely reflects the combined effects of high-resolution acquisition and deep learning-based reconstruction.

OBJECTIVES:This study aimed to investigate the predictive value of longitudinal changes in tumor morphology and body composition during neoadjuvant chemotherapy (NAC) for progression-free survival (PFS) following radical resection of adenocarcinoma of the esophagogastric junction (AEG). METHODS:This retrospective study included 258 AEG patients receiving NAC at three hospitals. Clinical and pathological data were collected. Tumor morphological and body composition parameters were quantitatively assessed on venous phase CT images at pre-treatment (Pre) and post-treatment (Post) time points. The two measurements were compared and the reduction rate (Δ%) was calculated. Multivariate Cox regression analysis was used to identify independent predictors of PFS in AEG patients and a nomogram model was developed. The incremental predictive value of tumor morphological and body composition parameters was evaluated using the concordance index (C-index), net reclassification improvement, and integrated discrimination improvement. The goodness-of-fit of models was assessed via the Akaike information criterion and χ2 likelihood ratio test. The performance of the nomogram was evaluated by the area under the time-dependent receiver operating characteristic (tdROC) curve, calibration curves, and decision curve analysis. High-risk and low-risk subgroup analyses were performed according to nomogram scores. RESULTS:Multivariate Cox regression analysis showed that ypTNM staging, Post-tumor volume, and Δ%-skeletal muscle index (SMI) were independent predictors of PFS. The nomogram incorporating these predictors demonstrated significantly superior discrimination over ypTNM staging alone in both the training cohort (C-index: 0.744; 95 % CI: 0.670-0.790; P = 0.004) and an external validation cohort (C-index: 0.738; 95 % CI: 0.615-0.807; P = 0.024). tdROC analysis showed that the nomogram achieved area under the curve (AUC) values of 0.815 and 0.791 for predicting 1- and 2-year PFS, respectively, in the training cohort. These findings were corroborated in the external validation cohort, with corresponding AUCs of 0.761 and 0.746 for 1- and 2-year PFS, respectively. Moreover, according to the score of the nomogram, patients can be effectively divided into low-risk and high-risk groups. CONCLUSION:The nomogram, incorporating ypTNM staging, Post-tumor volume, and Δ%-SMI, demonstrated robust performance in predicting PFS in AEG patients. This model significantly outperformed traditional ypTNM staging alone and may help guide personalized postoperative monitoring strategies.

A novel risk stratification model based on Lung-RADS® v2022 and CT features was constructed and validated for predicting invasive pure ground-glass nodules (pGGNs) in China. Five hundred and twenty-six patients with 572 pulmonary GGNs were prospectively enrolled and divided into training (n = 169) and validation (n = 403) sets. Utilising the Lung-RADS® v2022 framework and the types of GGN-vessel relationships (GVR), a complementary Lung-RADS® v2022 was established, and the pGGNs were reclassified from categories 2, 3 and 4x of Lung-RADS® v2022 into 2, 3, 4a, 4b, and 4x of cLung-RADS® v2022. The cutoff value of invasive pGGNs was defined as the cLung-RADS® v2022 4a-4x. Evaluation metrics like recall rate, precision, F1 score, accuracy, Matthews correlation coefficient (MCC), and the area under the receiver operating characteristic curve (AUC) were employed to assess the utility of the cLung-RADS® v2022. In the training set, compared with the Lung-RADS 1.0, the AUC of Lung-RADS® v2022 were decreased from 0.543 to 0.511 (p-value = 0.002), and compared to Lung-RADS 1.0 and Lung-RADS® v2022, the cLung-RADS® v2022 model exhibited the highest recall rate (94.9 https://www.medicalresearch.org.cn/search/research/researchView?id=a97e67d8-1ee6-40fb-aab1-e6238dbd8f29 .

Deep learning (DL) demonstrates high sensitivity but low specificity in lung cancer (LC) detection during CT screening, and the seven Tumor-associated antigens autoantibodies (7-TAAbs), known for its high specificity in LC, was employed to improve the DL’s specificity for the efficiency of LC screening in China. To develop and evaluate a risk model combining 7-TAAbs test and DL scores for diagnosing LC with pulmonary lesions < 70 mm. Four hundreds and six patients with 406 lesions were enrolled and assigned into training set (n = 313) and test set (n = 93) randomly. The malignant lesions were defined as those lesions with high malignant risks by DL or those with positive expression of 7-TAAbs panel. Model performance was assessed using the area under the receiver operating characteristic curves (AUC). In the training set, the AUCs for DL, 7-TAAbs, combined model (DL and 7-TAAbs) and combined model (DL or 7-TAAbs) were 0.771, 0.638, 0.606, 0.809 seperately. In the test set, the combined model (DL or 7-TAAbs) achieved achieved the highest sensitivity (82.6

Background To establish and validate a dual-energy CT (DECT) radiomics model for predicting neoadjuvant chemotherapy (NAC) response in locally advanced gastric cancer (LAGC) across two vendors. Methods This was a secondary analysis drawn from a prospective cohort using DECT data of patients undergoing NAC followed by gastrectomy. Patients were stratified as responders (TRG 0/1) or non-responders (TRG 2/3) based on tumor regression grade (TRG). Radiomics features were extracted from polychromatic images at arterial/venous/delayed phases for building CECT model; Radiomics features extracted from polychromatic images, monochromatic (40 keV, 100 keV) and iodine maps were used to construct DECT model. Predictive features were selected via the least absolute shrinkage and selection operator regression method in the training cohort and tested in the validation cohort. Performances of models were evaluated using areas under the receiver operating characteristic curves (AUCs). Results In total, 317 patients were recruited: 221 at training dataset (59.9 ± 9.7 years, 37 females, 184 males) and 96 at validation dataset (61.5 ± 8.0 years, 18 females, 78 males). No clinical factors were found to be related with TRG status. The DECT model outperformed CECT model in the training dataset (AUC: 0.806 vs. 0.729, p = 0.041) and showed non-significant superiority in the validation dataset (AUC: 0.752 vs. 0.679, p = 0.225). High-risk patients defined by DECT model had significantly worse overall survival (HR = 1.996, p = 0.012) and disease-free survival (HR = 1.873, p = 0.037) than low-risk counterparts. Conclusion DECT radiomics demonstrates favorable performance in predicting NAC response and stratifying survival outcomes in LAGC, with cross-vendor generalizability supporting potential clinical utility.

Background:Accurate identification of brain metastases is critical for radiotherapy planning. The aim of this study was to evaluate the effectiveness of contrast-enhanced T1-weighted fluid-attenuated inversion recovery (CE-T1FLAIR) sequences combined with the three-dimensional modulated flip-angle technique to enhance the visualization of metastatic tumors. Methods:A retrospective study of 326 patients who were pathologically diagnosed with malignant tumors and clinically suspected of brain metastases from October 2023 to February 2024 was conducted. Magnetic resonance images from the modulated flip-angle technique in refocused imaging with extended echo trains (MATRIX) CE-T1FLAIR, three-dimensional magnetization-prepared rapid gradient echo (3D GRE) fast-spin echo (FSE) with modulated flip-angle flow-sensitive preparation (fsp) CE-T1FLAIR, and 2D FSE CE-T1FLAIR sequences were analyzed by three independent radiologists in a double-blind manner. Detection rates were compared with the χ2 test or Fisher exact test, with multiple comparisons conducted via the Bonferroni method. Interrater reliability was assessed with the Kappa consistency test. Results:A total of 176 patients with 887 brain metastases were included based on the linked intelligence brain metastasis artificial intelligence-assisted detection system and clinical follow-up. The MATRIX CE-T1FLAIR sequence had an overall lesion detection rate of 98.9%, significantly higher than the 93.5% for 3D GRE fsp CE-T1FLAIR and 80.6% for 2D FSE CE-T1FLAIR. For lesions <5 mm, the detection rate was 98.8%, with significant differences between the sequences. Lesion distribution analysis indicated that MATRIX CE-T1FLAIR had a higher detection rate in the superficial area of the brain's convexity, gray-white matter junction, and basal ganglia. Conclusions:The MATRIX CE-T1FLAIR sequence demonstrates superior overall and subgroup detection rates, particularly for small brain metastases (<5 mm) in the superficial areas of the brain's convexity as compared to 3D GRE fsp CE-T1FLAIR and 2D FSE CE-T1FLAIR.

To investigate the diagnostic value of CE-MATRIX-T1FLAIR and 3D CE-T2FLAIR sequences based on Contrast Enhancement Modulated flip Angle Technique in Refocused Imaging with eXtended echo train (CE-MATRIX) technology for detecting Leptomeningeal Metastasis (LM) using Fluid Attenuated Inversion Recovery (FLAIR) imaging. This prospective study included 563 hospitalized patients with clinically suspected LM, diagnosed with malignant tumors between January 2022 and October 2023 at Henan Cancer Hospital. Both CE-MATRIX-T1FLAIR and 3D CE-T2FLAIR sequences were used for imaging. Two radiologists independently evaluated image quality, diagnostic confidence, and objective measurements, diagnosing LM as positive or negative, with disagreements resolved by consultation. Subjective and objective scores were compared using the Wilcoxon signed-rank test. The diagnostic performance of the sequences was compared using ROC curve analysis, with cerebrospinal fluid (CSF) cytology as the gold standard. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and area under the curve (AUC) values were calculated and compared using Z-tests. LM was confirmed in 321 patients. CE-MATRIX-T1FLAIR showed superior subjective scores in image quality and diagnostic confidence (p < 0.001). Though CE-MATRIX-T1FLAIR had a lower SNR (p = 0.013), it demonstrated higher sensitivity, specificity, PPV, NPV, accuracy, and AUC than 3D CE-T2FLAIR (p < 0.001). Both sequences provided effective diagnosis and differentiation of LM. CE-MATRIX-T1FLAIR offers superior diagnostic performance compared to 3D CE-T2FLAIR for LM, with slightly better subjective ratings despite a lower SNR. Both sequences are effective for diagnosing LM.

To explore the added value of the combination of radiomics and visual features based on contrasted enhanced computed tomography (CECT) images for predicting the invasiveness of pure ground-glass nodules (pGGNs). The clinical and imaging data of 123 patients with 143 pGGNs confirmed by surgical pathology were retrospectively analyzed. The lesions-based dataset was randomly divided with a ratio of 7:3 into training set and test set. Radiomics models and visual features model were constructed independently using logistic regression. Two combined model of 2D + and 3D + were also established. The performance of the five models was evaluated via receiver operating characteristic (ROC) curve analysis and the clinical validity was assessed by using the model’s integrated discrimination improvement (IDI) indices. The 3D + model and 2D + model performed better with higher AUC (training: 0.839/0.793; test: 0.829/0.794) than three independent models alone (all P < 0.05) and the DCA showed the IDI of 3D + model had a significant improvement in the test set than 2D + model, 3D radiomics model and visual features model (9.96