Search Results (110)

Objective: We evaluated whether a deep-learning model could predict the response to neoadjuvant chemotherapy (NAC) in breast cancer using the pre-treatment B-mode ultrasound. Methods: This retrospective study included 245 female patients (253 lesions) treated with NAC between 2017 and 2019. Lesions were categorized as complete response (CR; 103) or non-CR (150) based on postoperative pathology. We trained ResNet18-based models using pre-treatment B-mode ultrasound images (Image) and clinical features. Three training strategies were evaluated: training from scratch (SC); transfer learning (TL); and domain-specific pretraining (USP). Predictive performance was assessed using descriptive statistics. Results: The best-performing model (USP Image) achieved 0.76 accuracy (specificity: 0.80; sensitivity: 0.72), significantly outperforming all other models, including those that used additional clinical features ($p<0.05$). USP improved performance across most model types compared to SC and TL, highlighting the value of domain-specific pretraining. Clinical features added value with SC or TL, but not with USP, suggesting that pretrained models can extract the most relevant information directly from images. Grad-CAM analysis revealed that non-CR predictions focused on the tumor and posterior shadowing—features linked to chemoresistant subtypes. CR predictions focused mainly on more heterogeneous, peritumoral regions. Conclusion: This finding underscores ultrasound’s potential as a low-cost, accessible tool for predictive oncology in personalized, AI-driven treatment planning. Full article

Background/Objectives: To evaluate the incidence of rapid recurrence within 6 months of curative resection for hepatocellular carcinoma (HCC) and to identify preoperative gadoxetic-acid-enhanced MRI features associated with rapid recurrence (<6 months) in the entire cohort. Methods: This retrospective study included 200 patients who underwent curative resection for HCC and had preoperative gadoxetic-acid-enhanced MRI between January 2016 and December 2023. Patients were categorized into a rapid recurrence group (n = 21) and a non-rapid recurrence group (n = 179). Preoperative MRI features, including tumor size, multiplicity, tumor margin, arterial peritumoral enhancement, peritumoral hepatobiliary phase (HBP) hypointensity, diffusion restriction, apparent diffusion coefficient (ADC) values, and presence of non-hypervascular hepatobiliary phase hypointense nodule (NHHN), were evaluated. Univariate and multivariate Firth penalized logistic regression analyses were performed. Results: Rapid recurrence occurred in 10.5% (21/200) of patients (median, 4.0 months). Multivariate analysis revealed that larger tumor size (odds ratio [OR], 1.25 per 1-cm increase; p = 0.012) and the presence of NHHN (OR, 11.30; p < 0.001) were independent predictors of rapid recurrence. A nomogram incorporating these features demonstrated excellent discriminative performance, with a bootstrap-corrected area under the curve (AUC) of 0.864 (95% CI, 0.791–0.922). Conclusions: The presence of NHHN and larger tumor size on preoperative MRI were associated with rapid recurrence (<6 months) after curative resection for HCC. These findings may provide additional support for preoperative risk stratification and the planning of postoperative surveillance strategies. Full article

Background/Objectives: Perirenal fat (PRF) constitutes a critical yet understudied component of the tumor microenvironment in clear cell renal cell carcinoma (ccRCC). While radiomics enables non-invasive tissue characterization, whether PRF-derived features can simultaneously predict both TNM stage and histological grade, and whether optimal peritumoral distances differ between these distinct biological targets, remains unexplored in the literature. Methods: This multi-cohort retrospective study included 474 histopathologically confirmed ccRCC patients from three independent datasets (2007–2023). Automated nnU-Net segmentation delineated tumors and kidneys. Concentric PRF regions were systematically generated at 1–10 mm radial distances, yielding 18 distinct regions of interest. From each ROI, 1409 radiomic features were extracted using PyRadiomics. Sequential feature selection employed correlation filtering, SHAP-guided elimination, and LASSO regularization. Multiple machine learning classifiers underwent hyperparameter optimization with rigorous cross-cohort validation. Results: Systematic ROI screening revealed target-specific optimal distances: 4 mm PRF for TNM staging versus 10 mm PRF for histological grading. For staging, the integrated model (tumor + PRF radiomics + clinical variables) achieved AUC 0.829 (95% CI 0.781–0.877), sensitivity 80.2%, and specificity 67.8%. For grading, the combined model achieved AUC 0.780 (95% CI 0.598–0.962), sensitivity 79.7%, and specificity 63.3%, significantly outperforming all single-compartment models (DeLong p < 0.001). Conclusions: This study establishes that PRF radiomics enables accurate simultaneous non-invasive prediction of both TNM stage and histological grade in ccRCC. The novel discovery that optimal peritumoral distances differ substantially by prediction target (4 mm versus 10 mm) suggests distinct biological underpinnings for stage- and grade-related microenvironmental alterations, with important methodological implications for radiomic model development in oncology. Full article

Objective: This prospective study investigated the use of H-scan ultrasound (US) imaging as a novel component of a multiparametric radiomic analysis framework for characterizing human breast cancer response to neoadjuvant chemotherapy (NAC) before and early after treatment initiation. Methods: Thirty breast cancer patients scheduled for NAC were scanned using a clinical US system (Logiq E9, GE HealthCare) equipped with a 9L-D linear array transducer. Radiofrequency (RF) data was obtained at baseline (pre-NAC) and after 10% and 30% of the complete dose of chemotherapy. The RF data was analyzed by a bank of 256 frequency-shifted bandpass filters to form H-scan US frequency images. Grayscale texture features were extracted from both B-scan and H-scan US images. In addition, US attenuation coefficient and speckle statistics based on the Nakagami and Burr distributions were estimated from the RF data. Data classification of tumor and peri-tumoral regions was performed using a novel three-dimensional (3D) score map based on support vector machine (SVM) modeling. Unlike conventional classifiers that report only a single prediction score, a 3D score map provides a visual representation of the classifier decision space, enabling interpretation of class separation and treatment-induced shifts in multiparametric US measurements. Results: The dataset was split into 10 disjoint partitions (90% training, 10% testing) to compute area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and accuracy measures. Actual patient response to NAC was assessed at surgery and categorized as either pathologic complete response (pCR) or non-pCR. Multiparametric US and data classification results at pre-NAC found AUC values of 0.78 after using only tumor information (p < 0.01), which increased to 0.81 with inclusion of peri-tumoral information (p < 0.01). Significant differences in multiparametric US measures from both cancer response types was found after integration of patient data collected at 10% completion of the NAC regimen (i.e., first NAC cycle), yielding an improved AUC of 0.86 (p < 0.001). Conclusions: Multiparametric US imaging with radiomic features from both the tumor and peri-tumoral regions is a promising noninvasive approach for monitoring early breast cancer response to NAC. Full article

Background/Objectives: The role of pre-treatment breast magnetic resonance imaging (MRI) findings and apparent diffusion coefficient (ADC) values in predicting pathologic complete response (pCR) in breast cancer patients receiving neoadjuvant chemotherapy (NAC) has not yet been sufficiently clarified, especially in the context of molecular subtype differences. In this study, we questioned whether these imaging parameters were independent predictors of pCR. Methods: This study retrospectively explored MRI characteristics of 188 patients who underwent NAC from 2015 to 2023. The patients were divided into the pCR-positive and pCR-negative groups—the latter comprising patients with partial response (n = 61) and stable disease (n = 90)—and were classified into four molecular subtypes: Luminal A/B, HER2-enriched, and triple-negative breast cancer (TNBC). The MRI parameters included pre-chemotherapy T2-weighted signal characteristics, shape features, contrast kinetics, peritumoral edema, and ADC MIN/ADC MAX. Post-treatment ADC and ΔADC were the post-chemotherapy MRI parameters. Independent predictors were evaluated by logistic regression and discriminant performance by ROC analysis. Results: The overall pCR rate was 19.7%. In multivariate analysis, T2-weighted isointense signal (OR = 4.50), uniform tumor shape (OR = 12.83), HER2-enriched subtype (OR = 6.03), TNBC (OR = 5.15), ADC MIN (OR = 1.41), tumor size (OR = 1.28), and kinetic Type 3 pattern (OR = 3.21) were identified as independent predictors. Pre-treatment ADC MIN yielded an AUC of 0.724, while post-treatment ADC achieved 100% sensitivity and 96.7% specificity (AUC = 0.967). Conclusions: MRI morphology and ADC values may make a meaningful contribution to the prediction of pCR when evaluated in the context of molecular subtype. Post-treatment ADC demonstrated particularly strong discriminatory performance; however, external validation in multicenter cohorts is required before clinical implementation. Full article

Objective: The objective of this study was to develop and validate a predictive model for MTM-HCC by integrating preoperative ultrasound (US) and contrast-enhanced ultrasound (CEUS) features with relevant clinical characteristics. Methods: This retrospective study analyzed data from patients with histopathologically confirmed hepatocellular carcinoma who underwent preoperative CEUS examination at the Ultrasound Department of the Lanzhou University Second Hospital between December 2021 and March 2025. The study cohort comprised 45 patients diagnosed with MTM-HCC and 194 patients with non-MTM-HCC. Ultrasound and CEUS images were independently reviewed by two senior abdominal radiologists with extensive experience in hepatic imaging, ensuring objective feature assessment. Clinical variables and imaging characteristics were systematically compared between the two groups to identify distinguishing patterns. To evaluate the associations among clinical data, ultrasound-derived features, and MTM-HCC, univariate analyses were first performed, followed by multivariate logistic regression to construct and assess predictive models. Results: A total of 239 patients (mean age: 57.28 ± 9.60 years; 187 males and 52 females) were included in the analysis. Among them, 45 HCC patients (18.8%) were classified as MTM-HCC. Multivariate analysis identified four independent predictors: elevated alpha-fetoprotein (AFP ≥ 467 ng/mL) (OR = 8.5, 95% CI: 4.2–17.30; p < 0.001), presence of non-enhancing necrotic areas (OR = 5.92, 95% CI: 1.82–19.30, p = 0.003), intratumoral arteries (OR = 6.61, 95% CI: 2.28–19.22, p < 0.001), and peritumoral feeding arteries (OR = 3.13, 95% CI: 1.15–8.50, p = 0.025). Conclusions: An integrated prediction model that combines ultrasound imaging and clinical parameters offers a feasible, non-invasive approach for accurate preoperative identification of MTM-HCC. Full article

Background/Objectives: Glypican-3 (GPC3), a membrane-bound heparan sulfate proteoglycan, has been identified as a promising target for both the diagnosis and treatment of hepatocellular carcinoma (HCC). However, the diagnosis of GPC3 expression mainly depended on invasive procedures. This study aimed to investigate the potential of dual-energy computed tomography (DECT)-derived parameters for noninvasive prediction of GPC3 expression in HCC. Methods: This retrospective study included 79 HCC patients with confirmed GPC3 immunohistochemistry and pretreatment contrast-enhanced DECT. Qualitative imaging features and quantitative DECT parameters, including iodine density of HCC (ID_Ca), normalized iodine density (NID), slope of spectral attenuation curve (λ_HU), and effective atomic number (Z_eff), were evaluated in both arterial and portal venous phases. Univariate and multivariate logistic regression analyses were employed to identify independent predictors, and a combined model was subsequently constructed. Receiver operating characteristic (ROC) curve analysis was performed to assess the diagnostic efficiency of imaging parameters in predicting GPC3 expression. Interobserver agreement of DECT parameters was evaluated using intraclass correlation coefficients (ICC). Results: GPC3-positive HCCs demonstrated significantly higher arterial phase (AP) ID_Ca, NID, λ_HU, and Z_eff (all p ≤ 0.001) than GPC3-negative HCCs. Multivariate logistic regression analysis identified NID-AP [Odds ratio (OR) = 2.00, p = 0.010] and peritumoral enhancement (OR = 9.25, p = 0.046) as independent predictors. The model combining NID-AP and peritumoral enhancement achieved the best diagnostic performance (AUC = 0.781, sensitivity = 67.86%, specificity = 78.26%) for predicting GPC3 expression. All DECT-derived parameters showed excellent interobserver reproducibility (ICC > 0.75 for all). Conclusions: Parameters derived from DECT, especially combining NID-AP and peritumoral enhancement, may be a potential tool to noninvasively predict GPC3 expression in HCC. Full article

Background/Objectives: Spread through air spaces (STAS) represents an aggressive invasion pattern in lung cancer and is associated with unfavorable oncologic outcomes. As STAS is currently identifiable only on postoperative pathology, reliable preoperative, noninvasive prediction remains a clinical challenge. This study aimed to evaluate the feasibility of predicting STAS using ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT)-derived radiomic and clinicoradiomic models. Methods: In this retrospective study, patients who underwent surgical resection for lung cancer with available preoperative ¹⁸F-FDG PET/CT imaging were analyzed. Radiomic features were extracted from intratumoral and peritumoral regions. Clinical, radiomic-only, and combined clinicoradiomic models were developed using LASSO-based feature selection and multivariable logistic regression. Model performance was evaluated using nested cross-validation, receiver operating characteristic analysis, calibration assessment, and decision curve analysis. Results: Radiomic features reflecting intratumoral metabolic characteristics and peritumoral tissue heterogeneity were significantly associated with STAS. The combined clinicoradiomic model demonstrated superior discriminative performance compared with the clinical and radiomic-only models (mean AUC ≈ 0.75), along with favorable calibration (Brier score = 0.20) and improved clinical net benefit across relevant threshold probabilities. Lower eosinophil count, lower SUVmin_tumor, and lower intratumoral SUV skewness emerged as independent predictors of STAS. Conclusions: Preoperative prediction of STAS in lung cancer is feasible using PET/CT-based radiomic analysis integrating intratumoral, peritumoral, and clinical features. This noninvasive approach provides biologically relevant information beyond conventional anatomical assessment and warrants further validation in prospective, multicenter cohorts. Full article

This study aims to investigate the predictive value of pre-treatment multi-phasic contrast-enhanced computed tomography (CECT) radiomic features for treatment resistance in patients with rat sarcoma virus (RAS)-mutated colorectal liver metastases (CRLMs) receiving bevacizumab-based chemotherapy. Seventy-three samples with RAS-mutated CRLMs receiving bevacizumab-combined chemotherapy regimens were evaluated. Radiomic features were extracted from arterial phase (AP), portal venous phase (PVP), AP-PVP subtraction image, and Delta phase (DeltaP, calculated as AP-to-PVP ratio) images. Three groups of radiomics features were extracted for each phase, including peritumor, core tumor, and whole-tumor regions. For each of the four phases, a two-sided independent Mann–Whitney U test with the Bonferroni correction and K-means clustering was applied to the remnant features for each phase. Subsequently, the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm was then applied for further feature selection. Six machine learning algorithms were then used for model development and validated on the independent testing cohort. Results showed peritumoral radiomic features and features derived from Laplacian of Gaussian (LoG) filtered images were dominant in all the compared machine learning algorithms; NB models yielded the best-performing prediction (Avg. training AUC: 0.731, Avg. testing AUC: 0.717) when combining all features from different phases of CECT images. This study demonstrates that peritumoral radiomic features and LoG-filtered pre-treatment multi-phasic CECT images were more predictive of treatment response to bevacizumab-based chemotherapy in RAS-mutated CRLMs compared to core tumor features. Full article

This study aimed to develop machine learning models based on conventional MRI features to classify HER2 expression levels in invasive breast cancer and explore their association with disease-free survival (DFS). A total of 678 patients from two centers were included, with Center 1 divided into training and internal test sets and Center 2 serving as an external test set. Random Forest models were trained to distinguish HER2-positive vs. HER2-negative (Task 1) and HER2-low vs. HER2-zero tumors (Task 2) using BI-RADS–based MRI features. SHapley Additive exPlanations were applied to rank feature importance, assist feature selection, and enhance model interpretability. DFS was analyzed using Kaplan–Meier curves and log-rank tests. In Task 1, key features included tumor size, axillary lymph nodes, fibroglandular tissue, peritumoral edema, and multifocal, achieving AUCs of 0.75 and 0.73 in the internal and external test sets, respectively. In Task 2, tumor size, peritumoral edema, and multifocal yielded AUCs of 0.73 and 0.72, respectively. Higher task-specific model scores were associated with shorter DFS in Task 1 (p = 0.037) and longer DFS in Task 2 (p = 0.046). MRI-based machine learning models can noninvasively stratify HER2 expression levels, with potential for prognostic stratification and clinical application. Full article

Background: Pancreatic and periampullary cancers are common tumors of the digestive tract. As a radical surgical approach, laparoscopic pancreaticoduodenectomy requires crucial preoperative assessment of its surgical difficulty. Materials and methods: A retrospective cohort of 150 patients who underwent LPD between June 2019 and June 2023 was enrolled. The criteria for defining the difficult group were identified as unplanned conversion to open procedure, intraoperative blood loss, and operative time. Participants were randomly allocated to a training set (n = 105) or a testing set (n = 45) in a 7:3 ratio. Hand-crafted radiomics (HCR) features and deep learning-derived radiomics (DLR) features were extracted from portal venous phase CT images, focusing on gross tumor volume and gross peri-tumor volume. A hybrid prediction model was developed using a support vector machine algorithm, with performance evaluated through receiver operating characteristic analysis, calibration curves, and decision curve analysis (DCA). Results: The combined model demonstrated significantly superior discriminative ability, achieving an area under the curve (AUC) of 0.942 (95% CI: 0.893–0.992) in the training set and 0.848 (95% CI: 0.738–0.958) in the testing set. This performance exceeded both the standalone HCR model (testing AUC = 0.754) and the DLR model (testing AUC = 0.816). DCA further confirmed the clinical utility of the combined model, showing the highest net benefit across threshold probabilities exceeding 20%. Conclusions: The novel integrated model combining hand-crafted and deep learning-derived radiomics features enables effective prediction of surgical difficulty in laparoscopic pancreaticoduodenectomy. Full article

Background:The presence of visceral pleural invasion (VPI) is associated with increased risk of recurrence and reduced overall survival following surgical resection. We aimed to develop machine learning (ML)-based classification models that integrate clinical variables and both tumoral and peritumoral radiomic features to predict VPI in patients with lung adenocarcinoma before surgery. Methods: We retrospectively enrolled 118 patients, including 80 (68%) without VPI and 38 (32%) with histologically confirmed VPI. All patients underwent preoperative contrast-enhanced CT scans. Tumor volumes were manually segmented, and isotropic expansions of 3, 5, and 10 mm were automatically generated to define peritumoral regions. The dataset was randomly split into training (70%) and validation (30%) cohorts. Radiomic features and clinical data were used to train multiple ML algorithms. Results: Pleural Tag Sign and the Worst Histotype were identified as the strongest clinical predictors of VPI. The combined model, integrating radiomics from the lesion and clinical variables, achieved the highest training accuracy of 0.88 (95% CI: 0.80–0.92) and validation accuracy of 0.83 (95% CI: 0.68–0.92). Conclusions: VPI is associated with detectable alterations in both tumoral and peritumoral microenvironment on contrast-enhanced CT. Incorporating radiomic features with clinical data enabled improved model performance compared to clinical-only models, yielding very good accuracies. This approach may support surgical planning and patient risk stratification. Further prospective studies are needed to validate these findings and assess their clinical impact. Full article

Background: Laryngeal and hypopharyngeal cancers (LHCs) exhibit heterogeneous outcomes after definitive radiotherapy (RT). Large language models (LLMs) may enhance prognostic stratification by integrating complex clinical and imaging data. This study validated two pre-trained LLMs—GPT-4o-2024-08-06 and Gemma-2-27b-it—for outcome prediction in LHC. Methods: Ninety-two patients with non-metastatic LHC treated with definitive (chemo)radiotherapy at Linkou Chang Gung Memorial Hospital (2006–2013) were retrospectively analyzed. First-order and 3D radiomic features were extracted from intra- and peritumoral regions on pre- and mid-RT CT scans. LLMs were prompted with clinical variables, radiotherapy notes, and radiomic features to classify patients as high- or low-risk for death, recurrence, and distant metastasis. Model performance was assessed using sensitivity, specificity, AUC, Kaplan–Meier survival analysis, and McNemar tests. Results: Integration of radiomic features significantly improved prognostic discrimination over clinical/RT plan data alone for both LLMs. For death prediction, pre-RT radiomics were the most predictive: GPT-4o achieved a peak AUC of 0.730 using intratumoral features, while Gemma-2-27b reached 0.736 using peritumoral features. For recurrence prediction, mid-RT peritumoral features yielded optimal performance (AUC = 0.703 for GPT-4o; AUC = 0.709 for Gemma-2-27b). Kaplan–Meier analyses confirmed statistically significant separation of risk groups: pre-RT intra- and peritumoral features for overall survival (for both GPT-4o and Gemma-2-27b, p < 0.05), and mid-RT peritumoral features for recurrence-free survival (p = 0.028 for GPT-4o; p = 0.017 for Gemma-2-27b). McNemar tests revealed no significant performance difference between the two LLMs when augmented with radiomics (all p > 0.05), indicating that the open-source model achieved comparable accuracy to its proprietary counterpart. Both models generated clinically coherent, patient-specific rationales explaining risk assignments, enhancing interpretability and clinical trust. Conclusions: This external validation demonstrates that pre-trained LLMs can serve as accurate, interpretable, and multimodal prognostic engines for LHC. Pre-RT radiomic features are critical for predicting mortality and metastasis, while mid-RT peritumoral features uniquely inform recurrence risk. The comparable performance of the open-source Gemma-2-27b-it model suggests a scalable, cost-effective, and privacy-preserving pathway for the integration of LLM-based tools into precision radiation oncology workflows to enhance risk stratification and therapeutic personalization. Full article

Background: Adamantinoma is a very rare primary malignant bone tumor. A histopathological grading is still lacking, and as a result, metastatic risk stratification at diagnosis is challenging. Due to this, imaging could play a role in prognosis prediction and treatment strategy assessment. We aimed to evaluate baseline imaging features and their correlation with the development of metastatic disease. Methods: We retrospectively collected clinical (metastatic disease) and radiological data at baseline (Conventional Radiography, CT, MRI) of all consecutive patients with a histopathological diagnosis of adamantinoma at our sarcoma center between 2006 and 2022. Tumor location, dimensions, main radiological pattern (lytic, sclerotic, mixed), Lodwick–Madewell grading, periosteal reaction, multifocality, soft-tissue extraskeletal component, peritumoral edema, peritumoral enhancement, and vascular invasion were analyzed. Associations between the above-mentioned radiological features and metastatic disease at diagnosis or during follow-up were assessed. Results: Twenty-two patients were included (15 [68.2%] women, median age 27 years old, range 7–58 years old). Six out of twenty-two patients (27.3%) developed distant metastases (only two of them were dedifferentiated adamantinoma): two patients (9%) presented with metastatic disease at diagnosis, while four patients developed metastases during follow-up (18.2%). The following radiological features represent a significant risk for metastatic disease (p = 0.01): (i) presence of an extra-skeletal component (Odds Ratio [OR] = 75.40; 95% CI = 3.15–1802.71), (ii) vascular invasion (OR = 121.00; 95% CI = 4.28–3424.73), (iii) diffuse peritumoral edema (OR = 75.40; 95% CI = 3.15–1802.71), (iv) peritumoral enhancement (OR = 84.33; 95% CI = 2.93–2423.26). All other features analyzed were not significantly associated with the onset of distant metastases. Based on these above-mentioned MRI features, we built two risk models for metastatic disease (excluding peritumoral enhancement, which was not available in five patients, to be applicable on unenhanced MRIs): Model (A) = simultaneous presence of two of those three features (2/3) with a sensitivity of 100% (54.07–100%) and a specificity of 93.75% (69.67–99.84%). Model (B) = simultaneous presence of all three features (3/3) with a sensitivity of 83.33% (35.88–99.58%) and a specificity of 100% (74.1–100%). Conclusions: An accurate evaluation of baseline imaging studies (particularly MRI) in patients affected by adamantinoma may significantly aid in prognosis prediction and the selection of high-metastatic-risk patients. For these patients, strict follow-up controls and more aggressive treatments should be suggested after multidisciplinary discussions in sarcoma centers. Full article

Background/Objectives: To identify preoperative MRI features using computer-assisted diagnosis (CAD) that are associated with invasive disease-free survival (IDFS) and distant metastasis-free survival (DDFS) in patients with primarily operable triple-negative breast cancer (TNBC). Methods: This retrospective study was approved by the institutional review board with informed consent was waived. Between January 2012 and December 2014, 74 consecutive women with primary TNBC (mean age, 51 years; range, 29–77 years) who underwent preoperative MRI were included and followed until August 2021. Dynamic contrast-enhanced and T2-weighted images were obtained using 3T scanners. Peritumoral edema and central necrosis were evaluated retrospectively. CAD was used to extract 3D diameters, angiovolume, and kinetic parameters, and kinetic heterogeneity was calculated. Cox proportional hazards models were used to assess associations between MRI features and IDFS and DDFS, adjusting for clinicopathologic factors. Results: During a median follow-up of 80.9 months, 12 patients developed invasive disease, and 8 developed distant metastasis. In multivariable analysis, peak enhancement (hazard ratio [HR], 1.40; 95% confidence interval [CI], 1.06–1.84; p = 0.019) and angiovolume (HR, 2.86; 95% CI, 1.26–6.47; p = 0.012) were independently associated with IDFS, whereas angiovolume (HR, 2.47; 95% CI: 1.28–4.78; p = 0.007) was independently associated with DDFS. Conclusions: Preoperative CAD-derived MRI features, particularly peak enhancement and angiovolume, were associated with IDFS in TNBC patients whereas angiovolume alone was associated with DDFS. Full article