Brain invasion serves as an independent diagnostic criterion for the diagnosis of WHO grade 2 meningiomas. This study aimed to evaluate the value of radiomic models constructed with features extracted from different regions of the brain-to-tumor interface to predict potential brain invasion of meningiomas. The cohort of this retrospective multicenter study included 915 consecutive patients from three centers with histopathologically confirmed meningiomas. Brain invasion status was determined separately using predefined criteria based primarily on histopathological findings, with operative records used as supplementary information. Nine radiomic models were developed using radiomic features extracted from nine different regions of interest: 3-5 mm regions of the brain-to-tumor interface (BTI), brain-to-tumor inner interface (BTII), and brain-to-tumor outer interface (BTOI). The features were derived from T1W, T2-FLAIR, and CE-T1W images. A logistic regression classifier was trained for model development. Model discrimination was evaluated via receiver operating characteristic (ROC) analysis, while clinical utility and incremental predictive value were quantified using decision curve analysis (DCA) and the integrated discrimination improvement (IDI) index. As compared to models constructed with features from the 3-5 mm BTII and BTI regions, the BTOI radiomic models demonstrated enhanced predictive capability for meningioma brain invasion. The area under the curve (AUC) values for the 3-, 4-, and 5-mm BTOI models were 0.845, 0.845, and 0.905 with the test cohort, and 0.818, 0.825, and 0.874 with the external validation cohort, respectively. DCA confirmed the clinical superiority of the BTOI models, which provided the highest net benefit across various threshold probabilities. Moreover, the IDI index indicated that the BTOI features offered statistically significant incremental value for invasion risk stratification as compared to the BTII and BTI models (p < 0.05). The BTOI radiomic models demonstrated superior performance in predicting brain invasion in meningiomas compared with models constructed from the BTI and BTII regions. These findings suggest that the peritumoral brain tissue adjacent to the tumor may contain particularly informative imaging characteristics related to invasion and provide imaging-based support for optimizing ROI delineation.
This study aimed to develop a dynamic nomogram in which clinical indicators are integrated with magnetic resonance imaging (MRI) radiomics to predict axillary lymph node metastasis (ALNM) in patients with breast cancer. Our retrospective study included 221 pathologically confirmed patients with breast cancer. Radiomic features were extracted from dynamic contrast-enhanced MRI (DCE-MRI) and fat-suppressed T2-weighted imaging (FS-T2WI) datasets. After feature screening, a support vector machine (SVM) algorithm was employed to establish radiomic models and calculate radiomic scores. Clinical independent predictors were identified through univariate and multivariate logistic regression analyses. A nomogram was established on the basis of the radiomic scores obtained with the optimal SVM model and clinically independent predictors, and it was subsequently transformed into a dynamic nomogram. Model performance was evaluated by the receiver operating characteristic curve and area under the curve (AUC). Shapley additive explanation (SHAP) was applied to interpret the contribution of clinical predictors. Calibration and decision curves were employed to assess nomogram performance. The platelet-to-lymphocyte ratio (PLR), breast imaging reporting and data system (BI-RADS) classification, and ALN status on MRI were identified as independent predictors of ALNM (all p < 0.05). SHAP analysis identified PLR as the top contributor. The clinical model developed with the three predictors achieved AUCs of 0.845 and 0.706 in the training and validation cohorts, respectively. Among the SVM models, the DCE-MRI and FS-T2WI fusion sequence model outperformed the single-sequence models, with AUCs of 0.868 and 0.875, respectively, in the training and validation cohorts. When clinical predictors were incorporated into the fusion sequence model, the nomogram achieved AUC values of 0.930 and 0.928 in the training and validation cohorts. Decision curve analysis demonstrated that the nomogram has significant clinical value. A nomogram model integrating MRI-derived radiomic scores, BI-RADS classification, ALN status, and PLR demonstrated good performance in predicting ALNM in patients with breast cancer.
WB-MRI is a radiation-free imaging technique widely used in oncology. Despite clinical advantages, WB-MRI can cause significant patient anxiety and claustrophobia, affecting scan completion and patient adherence. Non-invasive interventions to reduce MRI-related distress are still limited in clinical practice. This study evaluates the use of audiovisual and auditory entertainment for reducing anxiety during WB-MRI. In this prospective, quasi-experimental trial, 213 cancer patients (median age: 52.0[44-60]; female 70%) undergoing WB-MRI were equally assigned to Video (nature landscapes and sounds), Music (relaxing instrumental), or Standard (no distraction) groups. Anxiety was assessed using the State Trait Anxiety Inventory (STAI) before and after the exam. Additional measures included the Distress Thermometer, WB-MRI acceptance questionnaire, and claustrophobia ratings. Data analysis compared anxiety changes, interactions between baseline anxiety and claustrophobia, and logistic regression for anxiety reduction likelihood. Baseline anxiety was comparable across groups. Anxiety reduction was greater in the Video (mean STAI = -5.57) and Music (-2.48) groups compared to the Standard (0.33) group, with the Video group significantly outperforming the Music group (p = 0.006). Logistic regression showed patients in Video (95%CI:3.77-68.22, p < 0.001) and Music (95%CI:1.47-29.8, p = 0.02) groups were significantly more likely to reduce anxiety than controls. Patients with severe claustrophobia and high baseline STAI exhibited pronounced improvements, particularly with Video (β = - 0.38, p < 0.001). The alluvial plot shows that the Video and Music groups experienced a marked reduction in the proportion of patients with severe anxiety, while this proportion slightly increased in the standard condition after the exam. Audiovisual and auditory entertainment significantly reduced anxiety during WB-MRI, with audiovisual showing the greatest benefit. The simultaneous involvement of multisensory distraction enhances tolerance to procedures, especially in highly anxious and claustrophobic patients. These findings advocate for the adoption of a patient-centered audiovisual system as a non-invasive, practical strategy to enhance patient experience and increase the likelihood of successful scan completion in WB-MRI practice. The results hold significant implications for optimizing patient compliance, offering tangible clinical and economic benefits, including reduced sedation requirements, prevention of long-term phobic responses, minimized motion artifacts, and improved image quality. Retrospectively registered on ClinicalTrials.gov NCT06332001 [first posted March 26, 2024].
The aim of this study was to evaluate the visibility of colorectal liver metastases (CRLM) using photon-counting detector computed tomography (PCD-CT) and to determine the optimal virtual monoenergetic image (VMI) and iodine map reconstructions for improved contrast detection between metastases and surrounding liver parenchyma. A total of 117 patients with 227 CRLM (up to three measurements per patient) who underwent abdominal PCD-CT for staging between 09/2022 and 08/2024 were retrospectively included. VMI were reconstructed at energy levels between 40 and 90 keV (in 10 keV increments), and scanner-generated iodine maps were additionally analysed. To quantify contrast between CRLM and liver parenchyma, the parenchyma-to-lesion ratio (PLR) was calculated for each VMI and iodine map. The contrast-to-noise ratio (CNR) was determined based on attenuation values of the metastases and the bilateral musculus erector spinae, as well as its standard deviation. For the iodine map, lesion and parenchyma iodine concentrations were used analogously. Subjective assessment of metastases visibility on the three best VMIs in PLR and CNR (40-60 keV) and iodine maps were independently performed by three radiologists. Lesion and liver attenuation decreased steadily with higher keV levels. Iodine maps showed markedly higher iodine concentration in liver parenchyma than in metastases. The PLR was highest on the iodine map (3.29 ± 2.01), followed by 40 keV (2.19 ± 0.73). Regarding CNR, the 40 keV VMI showed the highest value (1.49 ± 1.70), followed by the iodine map (1.09 ± 0.99). CNR values decreased further at higher energies and significantly reduced at 70-90 keV. Paired superiority testing confirmed 40 keV as the best-performing VMI, showing significantly higher CNR than the iodine map, whereas PLR remained superior on the iodine map. Subjective ratings indicated that the 50 keV VMI provided the best visibility of CRLM. The iodine map consistently received lower subjective ratings across all criteria. Both iodine maps and low-keV VMIs, particularly at 40 keV, demonstrated high PLR and CNR values, contributing to improved depiction of CRLM in PCD-CT. The complementary use of these reconstructions may enhance lesion detection and overall diagnostic confidence.
This study aimed to develop and validate preoperative contrast-enhanced cone-beam breast CT (CE-CBBCT)-based radiomics, deep learning, and combined models for predicting Ki-67 expression status and to preliminarily explore their prognostic value, with the goal of informing preoperative treatment decisions. This two-centre study enrolled 636 breast cancer patients, divided into training (n = 310), internal validation (n = 77), external test (n = 78), and prognostic (n = 171) cohorts. Two Ki-67 cut-offs (14% and 30%) were applied. Radiomics models were constructed using five machine learning classifiers, and deep learning models using four three-dimensional architectures (ResNet50, ResNet101, DenseNet121, and ShuffleNet). The best-performing models from each category were integrated with the clinical model using logistic regression to establish combined models. Prognostic value was evaluated using disease-free survival (DFS) in the prognostic cohort. Among the radiomics models, those based on ExtraTrees and eXtreme Gradient Boosting performed best for the 14% and 30% cut-offs, respectively, with areas under the receiver operating characteristic curves (AUCs) of 0.885 and 0.831 in the external test cohort. Among the deep learning architectures, ShuffleNet performed best, with corresponding AUCs of 0.630 and 0.825. Combined models achieved the highest AUCs of 0.895 and 0.869 for the 14% and 30% cut-offs, respectively, in the external test cohort. However, they did not significantly outperform the corresponding radiomics models. In survival analysis, the radiomics score for predicting Ki-67 ≥ 30% was an independent prognostic factor for DFS (hazard ratio: 1.661; 95% confidence interval: 1.142-2.417; P = 0.008). CE-CBBCT-based radiomics models demonstrated utility as non-invasive tools for the preoperative prediction of high Ki-67 expression defined at both the 14% and 30% cut-offs. The prognostic value of the radiomics score for predicting Ki-67 ≥ 30% was preliminarily demonstrated.
PSMA PET/CT has transformed prostate cancer management, yet its diagnostic utility is not absolute. It fails to detect PSMA-suppressed disease variants, including treatment-induced neuroendocrine prostate cancer (t-NEPC) and dedifferentiated tumors, and is inherently blind to second primary malignancies (SPMs). This study evaluated the additional diagnostic yield and direct clinical impact of incorporating [¹⁸F]FDG PET/CT into a systematic, indication-driven dual-tracer protocol in high-risk prostate cancer patients. In this retrospective, STROBE-compliant, single-center cohort study, 82 patients with histologically confirmed prostate cancer who underwent both [⁶⁸Ga]Ga-PSMA-11 and [¹⁸F]FDG PET/CT were analyzed. Clinical indications encompassed suspected SPM, suspected dedifferentiation or t-NEPC, initial staging of very high-risk disease, pre-[¹⁷⁷Lu]Lu-PSMA therapy evaluation, and equivocal PSMA findings. The primary endpoint was the proportion of patients in whom [¹⁸F]FDG PET/CT provided additional diagnostic information. Secondary endpoints included concordance analysis and the rate of management changes. Median age was 72 years (IQR 65-75). [⁶⁸Ga]Ga-PSMA-11 demonstrated superior prostate cancer detection compared to [¹⁸F]FDG (70.7% vs. 25.6%; McNemar's p < 0.0001). However, [¹⁸F]FDG PET/CT provided clinically critical additional information and directly altered management in 51.2% of patients (95% CI: 40.6-61.7%). The most impactful contributions were histopathologically confirmed SPM detection (30.5%, predominantly lung and colorectal adenocarcinomas) and identification of PSMA-negative/FDG-positive discordant disease signaling dedifferentiation. No significant correlation was found between PSMA and FDG SUVmax values (Spearman rho = 0.149, p = 0.422), underscoring their biologically distinct and complementary targets. [¹⁸F]FDG PET/CT is not a redundant adjunct to PSMA imaging; it fills a critical diagnostic blind spot. When applied through an indication-driven framework, dual-tracer imaging directly reshapes clinical decision-making in more than half of selected high-risk prostate cancer patients, enabling detection of occult malignancies and guiding pivotal treatment decisions that PSMA PET/CT alone cannot support.
To investigate the use of contrast-enhanced mammography (CEM) for preoperative prediction of lymphovascular invasion (LVI) status in invasive breast cancer. A total of 243 female patients diagnosed with invasive breast cancer (median age: 49 years; range: 27-77 years) who received preoperative CEM examination in our hospital between September 2018 and February 2024 were retrospectively collected and analyzed. The study population were chronologically divided into training and test datasets in an approximate ratio of 7:3. LVI status was determined using postoperative histopathologic examination. CEM features were analyzed on the low energy and the recombined images. To identify independent predictors for LVI status, univariable and multivariable logistic regression analyses were performed on CEM and clinicopathologic features. Logistic regression and six machine learning methods were used to construct prediction models in the training dataset, and their performance were evaluated with ROC curve in the test dataset. In training and test datasets, the rates of LVI-positive were 39% (67 of 172) and 34% (24 of 71), respectively. High Ki67 index, BI-RADS category 5, breast composition category c/d, axillary adenopathy, mild to marked background parenchymal enhancement level, and lesion with complete enhancement or enhancement extending on CEM images were significantly correlated with LVI-positive (all P < 0.05) and were incorporated to construct prediction models. The AUCs of seven prediction models were in the range of 0.713-0.850 in the test datasets, where the logistic regression model yielded an AUC of 0.835 (95%CI: 0.717-0.924), showing similar or higher AUC than the six machine learning models. CEM could be useful for preoperative noninvasive prediction of LVI status in invasive breast cancer. The prediction model integrating contrast-enhanced mammography features and Ki67 index may serve as a complementary tool to assist clinicians in preoperative prediction of lymphovascular invasion status in patients with invasive breast cancer.
To construct a machine learning model using contrast-enhanced CT radiomics and clinical indicators, aiming to improve the diagnostic accuracy for lymph node metastasis in children with peripheral neuroblastoma and provide practical evidence for clinical diagnosis and treatment. A total of children with pathologically confirmed neuroblastoma were retrospectively enrolled between February 2014 and December 2024, and then randomly divided into a training set and a test set via random sampling. Radiomics features were extracted separately from CT images of the arterial phase and venous phase. In the training set, four radiomics models, one clinical model, and one combined model incorporating radiomics and clinical features were constructed respectively using the filtered radiomics features and clinical features. All models were validated against the pathological reference standard in the test set, and the area under the receiver operating characteristic curve of each model was calculated. The clinical utility of each model was evaluated using the decision curve analysis curve. The optimal model was visualized with a nomogram, and the diagnostic gain of the nomogram for evaluating lymph node metastasis in neuroblastoma children was quantified via human-machine comparison. A total of 225 children with neuroblastoma were enrolled in this study, (with a mean age of 2.23 ± 2.34 years and an age range of 0-13 years). All subjects were randomly divided into a training set (n = 157) and a test set (n = 68) at a ratio of 7:3. Compared with four radiomics models (Arterial phase, Venous phase, Delta-Absolute, Delta-Relative) and one clinical model (Ki-67), the nomogram integrating radiomics and clinical features (Arterial phase + Delta-Relative + Ki-67) exhibited superior diagnostic performance in evaluating lymph node metastasis in pediatric peripheral neuroblastoma. The AUC values of the nomogram reached 0.937 and 0.829 in the training set and validation set, respectively. In the human-machine comparison experiment, the diagnostic accuracy of radiologists for lymph node metastasis in neuroblastoma children was improved by 21% when assisted by the nomogram. The nomogram combining contrast-enhanced CT radiomics and clinical indicators has significant diagnostic value in evaluating lymph node metastasis in pediatric patients with peripheral neuroblastoma. Moreover, it can substantially improve the diagnostic accuracy of radiologists with different levels of clinical experience.
This study evaluates the utility of quantitative imaging biomarkers derived from whole-body diffusion-weighted MRI (WB-DWMRI) and [68Ga]GaPSMA-PET/CT in predicting lesion-level response to [177Lu]LuPSMA therapy in metastatic castration-resistant prostate cancer (mCRPC). Twenty-two patients with mCRPC who underwent WB-DWMRI and [68Ga]GaPSMA-PET/CT within three months before [177Lu]LuPSMA therapy were identified. The PSMA SUV (SUVmean, SUVSD, SUVpeak) and corresponding diffusion weighted imaging (DWI) parameters (ADCmean, ADCkurtosis, ADCvol) were extracted from five hottest lesions (highest SUVmean) on PSMA-PET/CT. Lesion response was assessed using modified PERCIST and MET-RADS-P criteria. Multilevel logistic regression and area under receiver operating characteristic (AUROC) analyses identified predictive biomarkers. 65 bone lesions and 30 lymph nodes were analysed pre- and post-therapy. Forty-four (68%) bone and 16 (53%) lymph nodes lesions responded to treatment. SUVmean and SUVpeak were almost identical (rank-correlation = 0.97) and had high predictive performance for response with AUROC of 0.74 (95% CI: 0.62-0.86) and 0.75 (95% CI: 0.61-0.88) respectively. Lesion volume showed good performance (AUROC = 0.69, 95% CI 0.57-0.82) and moderately correlated with SUVmean (rank-correlation = 0.43). Neither ADCmean (AUROC = 0.45, p = 0.47) or ADCkurtosis (AUROC = 0.49, p = 0.171) were predictive. On regression modelling, a 1-unit volume increase, raised response odds by 4.3 (95% CI 0.7-27) in bone lesions, and 6.2 (95% CI 0.6-67) in lymph nodes (p = 0.06). A 1-unit SUVmean increase raised response odds by 1.5 (95% CI 1.1-2.0) in bone lesions and 1.2 (95% CI 1.0-1.4) in lymph nodes (p < 0.01). No evidence suggested combining volume with SUVmean enhanced predictive performance over SUVmean alone (p = 0.58). Baseline PSMA SUVmean, SUVpeak, and volume are promising predictive biomarkers for lesion-level response to [177Lu]LuPSMA therapy. Combining functional and structural imaging biomarkers could improve treatment stratification and response assessment. Further validation in larger studies, alongside patient-level analysis, and expansion beyond the five lesions is needed to refine predictive models for clinical application.
Lymphovascular invasion (LVI) signifies poor prognosis in bladder cancer, yet reliable preoperative prediction remains challenging, limiting personalized treatment planning. In this multi-center retrospective study, 543 bladder cancer patients were enrolled. Of these, 473 patients from two centers were randomly split into training and internal test sets at an 8:2 ratio, while 70 patients from six additional centers constituted an independent external test set. After tumor and peritumoral segmentation, a hybrid model that integrates deep learning and radiomics features was developed. The hybrid model was compared against ten baseline models, including image-only and radiomics-only deep learning models, as well as radiomics-based machine learning approaches. Model performance was assessed using the area under the receiver operating characteristic curve (AUC) with five-fold cross-validation. Prognostic value was evaluated by Kaplan-Meier survival analysis with the log-rank test. The hybrid model achieved AUCs of 0.77 (internal test) and 0.75 (external test), surpassing all baseline models. The model-predicted LVI status significantly stratified overall survival in the training set (p < 0.001) and the internal test set (p = 0.003), with a non-significant trend observed in the external test set (p = 0.151). The proposed non-invasive hybrid model can accurately predict LVI status preoperatively and demonstrates prognostic potential, thereby aiding risk stratification in bladder cancer.
This study aimed to evaluate the feasibility of employing habitat-based radiomic distributions in ultrasound (US) images to quantitatively characterize intratumoral heterogeneity. It also explored the potential of this approach to predict lymphovascular invasion (LVI) in breast invasive ductal carcinoma (IDC) patients and to identify the optimal extent of multiple peritumoral regions. A total of 408 women diagnosed with IDC from January 2020 to October 2023 were enrolled in this retrospective cohort study from two medical centers. Intratumoral areas were partitioned into four distinct habitat areas using K-means cluster analysis, while peritumoral regions were expanded at increments of 2, 4, and 6 mm. Radiomic features were independently extracted from the intra- and peri-tumoral areas, and habitat subregions for developing predictive models. These models incorporated three machine learning classifiers: Random Forest, Extreme Gradient Boosting (XGBoost), and Support Vector Machine (SVM), respectively. We subsequently established an integrated model encompassing intra- and peri-tumoral areas, and habitat radiomic features, and clinicopathological factors. The model performance was assessed through receiver operating characteristic (ROC), calibration curves, and decision curve analysis (DCA). Finally, SHapley Additive exPlanations (SHAP) and nonogram were applied to enhance model interpretability. The Random Forest model exhibited superior performance in terms of the area under the curve (AUC) values of 0.861 (95% CI: 0.809-0.912), 0.832 (95% CI: 0.746-0.919), and 0.810 (95% CI: 0.684-0.935) for the training, validation, and test sets, separately. Additionally, the peri-2 mm model surpassed the performance of the other models (peri-4 mm, peri-6 mm) in LVI prediction. The integrated model, encompassing peri-2 mm features, clinicopathological factors, and habitat models, achieved robust predictive performance with AUC values of 0.940 (95% CI: 0.907-0.973), 0.924 (95% CI: 0.875-0.973), and 0.852 (95% CI: 0.732-0.972) for each respective set. The integrated model yields the improved predictive performance in predicting LVI status, and the model offer a reliable and feasible preoperative prediction method to enhance the clinical management and therapeutic planning for IDC patients.
To evaluate contrast-enhanced ultrasound (CEUS) with VEGFR2-targeted microbubbles for longitudinal monitoring of early treatment effects during combined anti-PD-L1/anti-CTLA-4 immunotherapy in a murine colorectal cancer model. Murine colorectal cancer allografts (CT26) were established subcutaneously in 29 female Balb/c mice (therapy n = 15; control n = 14). Baseline CEUS with VEGFR2-targeted microbubbles was performed on day 7. The therapy group received intraperitoneal anti-PD-L1 and anti-CTLA-4 antibodies between days 7-15; controls received sham treatment. Follow-up CEUS was performed on days 14 and 21. Tumor perfusion (WiAUC) and VEGFR2 binding (SI8min, SI10min) were quantified, and immunohistochemistry assessed CD8, Ki-67, TUNEL, CD31, and VEGFR2. At FU1, WiAUC was significantly lower in the therapy group compared with controls (4,029 ± 61 vs. 6,391 ± 412; p < 0.001), and remained significantly lower at FU2 (1,028 ± 27 vs. 2,049 ± 39; p < 0.001). VEGFR2-targeted microbubble binding was consistently lower under therapy. At FU1, SI8min was 407 ± 11 vs. 626 ± 13 and SI10min was 409 ± 10 vs. 634 ± 28 (both p < 0.001). At FU2, SI8min was 106 ± 6 vs. 207 ± 4 and SI10min was 107 ± 5 vs. 207 ± 4 (both p < 0.001). Immunohistochemistry confirmed higher apoptosis and tumor-infiltrating lymphocytes, and lower proliferation, microvascular density, and VEGFR2 expression in treated tumors (all p < 0.05). VEGFR2-targeted CEUS enabled longitudinal monitoring of early response to dual anti-PD-L1/anti-CTLA-4 immunotherapy in colorectal cancer. Treated tumors showed lower perfusion and VEGFR2-targeted binding, paralleled by reduced vessel density and VEGFR2 expression as well as increased CD8 infiltration and apoptosis. These findings support VEGFR2-targeted CEUS as a promising noninvasive imaging biomarker for monitoring early immunotherapy-associated vascular changes. [Image: see text]
This study investigates the relationship between histopathological (HP) features, immunohistochemical (IHC) markers, 18F- FDG PET/CT parameters, and machine learning algorithms in patients diagnosed with invasive breast carcinoma, no special type. 384 patients were included in the study. 18F- FDG PET/CT images after diagnosis of invasive breast carcinoma, before treatment, were analyzed for the PET metabolic parameters retrospectively. PET/CT images of patients who met the inclusion criteria and were retrospectively analysed. Metabolic parameters including SUVmax, SUVmean, SUVpeak, MTV (Metabolic Tumour Volume), and TLG (Total Lesion Glycolysis) were calculated, considering SUVmax at 42% of the threshold value in the primary lesion for the PET/CT imaging. None of the parameters fit into the normal distribution. Therefore, Mann-Whitney-U, Kruskal-Wallis, and Spearman correlation tests were applied. The machine learning algorithms were also used to describe any connection between the HP and IHC properties of the tumors. Significant correlations were observed between PET-derived parameters (SUVmax, SUVmean, SUVpeak, MTV, and TLG) and HP/IHC characteristics, including Nottingham grade, molecular subtypes, and Ki-67 proliferation index. Machine learning models (XGBoost, Random Forest, LightGBM, CatBoost) demonstrated moderate success in identifying high-risk pathological and phenotypic features, with SUVmean identified as the most influential predictor via SHAP analysis. The coefficient of variation revealed low to moderate variability across most performance metrics, with CV values predominantly between 0.01 and 0.09, indicating stable and reliable model outputs. These findings suggest that integrating 18 F-FDG PET/CT metabolic parameters and machine learning algorithms can enhance non-invasive risk stratification and guide personalized treatment strategies in breast cancer management.
The purpose of this study was to evaluate the role of SWI and DWI in predicting the relapsed/refractory primary central nervous system lymphoma (R/R PCNSL). Seventy-seven patients with histologically confirmed PCNSL were enrolled. Patients were divided into R/R group (n = 40) and Non-R/R group (n = 37), according to follow-up outcomes. Demographics, pathological indicators, conventional MRI characteristics, pre-chemotherapy intratumoral susceptibility signal (ITSS) grade, ADC parameters, including pre-chemotherapy relative minimum ADC (rADCmin-pre) and relative mean ADC (rADCmean-pre), post-chemotherapy relative minimum ADC (rADCmin-post) and relative mean ADC (rADCmean-post), as well as the change in relative minimum ADC (rADCmin-change) and relative mean ADC (rADCmean-change), were compared between the two groups. Receiver operating characteristic curves and logistic regression analysis were used to assess the predictive performance. Compared with Non-R/R PCNSL group, R/R PCNSL group showed significantly lower Ki-67 index (P = 0.004) but higher pre-chemotherapy ITSS grade (P < 0.001). The comparison of rADCmin-pre and rADCmean-pre between the two groups showed no significant differences (P > 0.05). However, among the 50 patients with available post-treatment ADC data, rADCmin-post, rADCmean-post, rADCmin-change and rADCmean-change in R/R group were significantly lower than in Non-R/R group (all P < 0.001). The predictive performance of rADCmin-post, rADCmean-post, rADCmin-change and rADCmean-change was comparable to that of pre-chemotherapy ITSS grade for R/R PCNSL (0.894 vs. 0.714, 0.867 vs. 0.714, 0.877 vs. 0.714, 0.814 vs. 0.714, all P > 0.05), but superior to that of Ki-67 index (0.894 vs. 0.665, 0.867 vs. 0.665, 0.877 vs. 0.665, 0.814 vs. 0.665, all P < 0.05). The pre-chemotherapy ITSS grade, rADCmin-post, rADCmean-post, rADCmin-change and rADCmean-change may serve as preferable imaging biomarkers for predicting R/R PCNSL, and compare favorably with Ki-67 index.
To explore the prognostic value of tumor special growth rate (TSGR) in patients with NPC and investigate the potential of diffusion-weighted imaging (DWI) habitat analysis as a surrogate marker. Patients who underwent two MRI scans before any type of treatment at two institutes were included. Radiologists manually delineated the volume of the primary tumor, and the doubling time and TSGR were calculated. Correlation between TSGR and long-term survival was evaluated using multivariate Cox analysis. Moreover, a radiologist delineated the tumor on b = 0 s/mm2 images during the first MRI scan. Each tumor was divided into two subregions based on the individual mean apparent diffusion coefficient value (ADCMean), and the ADC value (ADCdown [mm2/s] and ADCup [mm2/s]) of each subregion was calculated. ADC values for diagnosing TSGR and predicting survival were evaluated and validated using 5-fold cross-validation (iterations = 1000) and in an independent cohort. A total of 98 patients were included (median age: 51 years; 24 females; training cohort: 64). The median doubling time was 131.9 days. Patients in the rapid-growth group (TSGR ≥ 1.4%) had significantly poorer survival than those in the slow-growth group (TSGR < 1.4%) in both the training and validation cohorts (p ≤ 0.006). ADCdown was a valuable factor in predicting TSGR (area under the curve: 0.792). Notably, the addition of ADCdown further improved the predictive value of TNM stage (0.663-0.848 vs. 0.631-0.785). Patients with high ADCdown value may have the rapid-growth potential and were more likely to have a poor prognosis.
Since the introduction of the Vesical Imaging-Reporting and Data System (VI-RADS), MRI has become an important imaging modality in the management of patients with bladder cancer. Its excellent diagnostic performance for determining muscle invasion in bladder cancer has been supported by numerous prospective and retrospective studies. Nevertheless, there needs to be continued improvement of the diagnostic performance of VI-RADS and sustained efforts to address remaining unmet clinical needs within the field of bladder cancer. In this paper, we highlight several such areas, some of which are actively being investigated. These include (1) whether to use intravenous contrast media or not (multiparametric vs. biparametric MRI), (2) quantitative metrics to enhance assessment of muscle invasion, (3) anatomic locations that come with pitfalls in staging, (4) introduction of bladder MRI image quality, (5) neoadjuvant chemotherapy VI-RADS (nacVI-RADS) and other considerations needed in the treatment response assessment after systemic therapies, (6) need for wider adoption, training, and implementation, and (7) application of artificial intelligence.
Cancer cachexia is a multifactorial syndrome characterised by involuntary weight loss and muscle wasting, commonly seen in patients with advanced malignancies. Lean Body Mass (LBM) assessment is crucial for the early indentification and monitoring of cancer cachexia. Dual-energy X-ray absorptiometry (DEXA) is widely accepted as the gold standard for evaluating body composition. This study aims to develop a method for estimating LBM using whole-body low-dose CT acquired during PET/CT imaging. A cohort of patients enrolled in a prospective pilot study on cancer cachexia were included in the analysis. All patients underwent both DEXA and [18F]FDG PET/CT imaging. Lean body mass was assessed by calculating the Total Body Lean Mass ratio (TBLMr) from DEXA (TBLMrDEXA) and also whole-body low-dose CT (TBLMrCT). A total of 32 (13-cachectic, 19-non-cachectic) patients with advanced malignancies were included in the analysis. The TBLMrCT and TBLMrDEXA were highly correlated (R2 = 0.92). Bland-Altman plots indicated a scaling bias with increasing TBLMrDEXA values, with the non-cachectic group exhibiting a tighter distribution (0.06 ± 0.05) compared to the cachectic group (0.03 ± 0.11). TBLMr values were significantly different between cachectic and non-cachectic groups for both the DEXA-derived (p < 0.004) and low-dose CT-derived (p < 0.010) measurements. Whole-body low-dose CT obtained from routine [18F]FDG PET/CT imaging can provide an accurate estimate of LBM, showing high concordance with DEXA-derived measurements and clinical diagnosis of cachexia. This method offers a practical method for monitoring cancer cachexia in patients undergoing routine [18F]FDG PET/CT scans.
Although pathological complete response (pCR) is generally associated with favorable survival outcomes, 10-15% of patients who achieve pCR after neoadjuvant chemoimmunotherapy still relapse within 3 years. This study aimed to explore the prognostic value of clinical features and metabolic parameters derived from 18-fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG PET/CT, hereafter referred to as PET/CT). We retrospectively analyzed 121 non-small cell lung cancer (NSCLC) patients with confirmed pCR of both primary tumors and lymph nodes after receiving neoadjuvant chemoimmunotherapy at Peking University Cancer Hospital. Univariate and multivariate Cox regression were used to identify prognostic factors, including clinical features, baseline (pre-neoadjuvant) and second (post-neoadjuvant) metabolic parameters on PET/CT-maximum standardized uptake value (SUVmax), peak standardized uptake value (SUVpeak), metabolic tumor volume (MTV), and total lesion glycolysis (TLG)-as well as their percentage changes prior to surgery. During a median follow-up period of 26 months, 11 patients experienced recurrence or metastasis. The 1-, 2-, and 3-year disease-free survival (DFS) rates were 98%, 92%, and 88%, respectively. PET-defined N2 (HR = 10.32; 95%CI: 1.90-55.99; p = 0.007), squamous cell carcinoma histology (HR = 0.27; 95%CI: 0.07-0.97; p = 0.05), and baseline MTV (HR = 1.02; 95%CI: 1.01-1.03; p = 0.01) were all correlated with DFS among NSCLC patients attaining pCR. When metabolic parameters were dichotomized into high and low groups, those pCR patients with higher metabolic activity had significantly shorter DFS than those with lower activity did. The corresponding hazard ratios for baseline PET/CT parameters were as follows: SUVpeak, 10.17 (95% CI: 1.3-79.51); MTV, 4.26 (95% CI: 1.13-16.1); and TLG, 8.43 (95% CI: 1.08-65.9). Among NSCLC patients achieving pCR after neoadjuvant chemoimmunotherapy, multivariable Cox regression analyses revealed that elevated baseline MTV and PET/CT-defined N2 status were independently associated with poor DFS, while squamous histology was associated with more favorable DFS. These findings highlight that even among pCR patients, baseline metabolic, histologic, and nodal staging characteristics retain important prognostic value. This real-world retrospective trial was registered at ClinicalTrials.gov on April 12th, 2025 (NCT06926179).
Contrast-enhanced CT (CE-CT) and MRI (CE-MRI) are routine examinations for staging of pancreatic cancer (PC). This study aimed to directly compare the diagnostic performance of the novel [18F]AlF-NOTA-FAPI-04 PET/CT against these established modalities for initial clinical staging. A total of 127 patients with initial PC who had simultaneously undergone [18F]AlF-NOTA-FAPI-04 PET/CT, CE-CT and CE-MRI were enrolled in this retrospective study. Their clinical features and imaging data were evaluated and staged according to TNM classification. Histopathological findings or follow-up data or multidisciplinary team (MDT) discussion served as the reference standard for the final diagnosis. For primary tumor (T) staging, CE-CT and CE-MRI demonstrated equivalent performance and outperformed [18F]AlF-NOTA-FAPI-04 PET/CT across all T stages except T1c. However, the PET/CT modality achieved significantly higher sensitivity, accuracy, and negative predictive value for both nodal (N) and distant metastasis (M) staging (all p < 0.05). It changed the N stage in 26 patients, altering management in 5 (3.93%), and revised the M stage in 16 and 18 patients compared to CE-CT and CE-MRI, respectively, leading to major treatment changes in 15 (11.81%) and 17 (13.38%) patients. A SUVmax cutoff of 4.85 for lymph node metastasis provided moderate sensitivity (67.1%) and high specificity (93.7%). [18F]AlF-NOTA-FAPI-04 PET/CT outperforms CE-CT and CE-MRI in evaluating N and M staging except T staging in initial pancreatic cancer, leading to a more accurate clinical staging and a more rational therapeutic decision-making. [18F]AlF-NOTA-FAPI-04 PET/CT should be considered a useful diagnostic tool for initial staging of pancreatic cancer.
To investigate whether the diffusion kurtosis imaging (DKI) technique can identify tumour response to an anti-angiogenic therapy in a pilot study of patients with RAS-mutant colorectal liver metastases. This prospective imaging study enrolled 20 participants receiving Regorafenib treatment. A target metastasis > 2 cm in each patient was imaged before and at 15 days after treatment on a 1.5T MR scanner using a coronal, free-breathing DKI protocol. Data were motion-corrected and modelled using both the mono-exponential and DKI models. Median values derived from voxel-wise analysis of the whole delineated tumour were reported for three parameters [apparent diffusion coefficient (ADC, 10- 3 mm2/s), apparent kurtosis (K, a.u.) and kurtosis-corrected apparent diffusion (D, 10- 3 mm2/s)] before and after treatment. A 5-patient supplementary cohort was used to assess the repeatability of DKI parameters using the Bland-Altman analysis. Changes in pre- and post-treatment measurements of the three parameters were assessed using Wilcoxon signed-rank tests (P < 0.05 was considered significant). Diffusion weighted imaging (DWI) and DKI parameter correlations were evaluated with Spearman tests. Functional MR parameters were also compared against Response Evaluation Criteria In Solid Tumours v.1.1 (RECIST) evaluations. Significant treatment-induced changes across the cohort were observed for all parameters: K decrease (0.907 vs. 0.786, P < 0.01; 13.3%), D increase (1.256 vs. 1.386 × 10- 3 mm2/s, P < 0.001; 10.4%) and ADC increase (0.910 vs. 1.035 × 10- 3 mm2/s, P < 0.001; 13.7%). For both visits, Spearman correlation tests found a moderate negative correlation between K and D, (r=-0.70; P < 0.001 and r=-0.58; P < 0.01) and a strong positive correlation (r = 0.84 and 0.88; P < < 0.001) between ADC and D. The repeatability R was lowest for K (51%), followed by D (12%), and ADC (9.4%). When compared to RECIST v.1.1 evaluations, K identified no clinical responders, whilst D and ADC identified 7/12 and 11/12 responders. Both ADC and DKI parameters showed a significant early cohort change to the anti-angiogenic effects of Regorafenib treatment in RAS-mutant colorectal liver metastases. The ADC parameter performed better than the K parameter in identifying patients benefitting from the treatment. NCT03010722 clinicaltrials.gov; registration date 6th January 2015.