BACKGROUND:Amide proton transfer weighted (APTw) imaging has demonstrated extensive clinical applications in diagnosing, treating evaluating, and prognosis prediction of breast cancer. There is a pressing need to automatically segment breast lesions on APTw original images to facilitate downstream quantification, which is however challenging. PURPOSE:To build a segmentation model on the original images of APTw imaging sequence by leveraging the varying contrasts between breast lesions and their surrounding glandular and fat tissues displayed on the original images of APTw imaging at different frequency offsets. METHODS:This paper proposes a network with multiple tasks, including a breast lesion segmentation model (task I) incorporating multiple images at different frequencies with different contrasts between tumor and surrounding tissues, an automatic classification of pathological task (task II), and an APTw parameter map fitting (task III). RESULTS:Compared with these advanced segmentation methods such as U-Net, segment anything model (SAM), segment anything in medical images (Med-SAM), and transfomer for MRI brain tumor segmentation (TransBTS), our method achieves higher accuracy (ACC). Furthermore, the model's interpretability facilitates the evaluation of how maps with varying gray contrasts contribute to the segmentation. Moreover, improving the ACC of segmentation can be accomplished through tasks such as pathological classification and parametric map fitting. CONCLUSIONS:The pathological classification task and parameter fitting task could improve the ACC of segmentation.

Abstract Background To investigate the usefulness of radiomics analysis based on voxel-wise mapping of DCE-MRI time-intensity-curve (TIC) profiles in quantifying temporal and spatial hemodynamic heterogeneity. Methods From December 2018 to August 2022, 428 patients with 639 breast lesions were retrospectively enrolled. The TIC profile of each voxel within the manually segmented 3D lesion was categorized into 19 subtypes based on wash-in rate (nonenhanced, slow, medium, and fast), wash-out enhancement (persistent, plateau, and decline), and wash-out stability (steady and unsteady). Three feature sets were calculated separately, including composition ratio (type-19) and radiomics features (type-19-radiomics) of 19 TIC profile subtypes, and radiomics features based on third-phase DCE-MRI images (phase-3-radiomics). Using support vector machine, four models (type-19, type-19-radiomics, type-19-combined, and phase-3-radiomics) were constructed to distinguish benign and malignant breast lesions. Results In differentiating benign and malignant lesions, both cross-validation and independent testing showed that type-19-combined model significantly outperformed phase-3-radiomics model (AUC = 0.906 vs. 0.823, P < 0.001, AUC = 0.867 vs. 0.762, P = 0.026). However, in cross-validation and testing, no significant difference in performance was observed between phase-3-radiomics model and type-19 model (P = 0.577 and 0.085), between phase-3-radiomics model and type-19-radiomics model (P = 0.182 and 0.200), or between type-19-radiomics model and type-19 model (P = 0.073 and 0.454). Conclusions In addition to radiomics analysis based on a single phase DCE-MRI, radiomics analysis of voxel-wise DCE-MRI time-intensity-curve (TIC) profiles map enables quantifying temporal and spatial hemodynamic heterogeneity simultaneously, thereby aiding in the differentiation of breast lesions.

Background Recent studies have shown that there exists a significant correlation between oral microbiome and the occurrence of malignancies. However, the prognostic significance of oral microbiome for cancer patients remains unclear. The purpose of this meta-analysis is to evaluate the impact of oral microbiome on the survival of patients with malignant neoplasms. Methods We conducted a thorough literature search of PubMed, Embase, and Cochrane Library databases until September 2022. The hazard ratio (HR) with a corresponding 95% confidence interval (CI) was analyzed using Review Manager 5.4 software for survival outcomes, including overall survival (OS), disease-specific survival (DSS), progression-free survival (PFS), and disease-free survival (DFS). Results A total of 15 studies, covering 5191 samples with various types of cancers, were selected based on specified inclusion and exclusion criteria. In both univariate and multivariate analysis, patients with low diversity of the oral microbiome, or those with Fusobacterium-high/positive, or P. gingivalis positive in cancer tissue displayed poorer OS (univariate HR = 1.74; 95% CI 1.15–2.62; P = 0.009; multivariate HR = 1.56; 95% CI 1.07–2.27; P = 0.02), DSS (univariate HR = 2.06; 95% CI 1.50–2.84; P < 0.00001; multivariate HR = 1.80; 95% CI 1.48–2.20; P < 0.00001), and PFS/DFS (univariate HR = 2.00; 95% CI 1.12–3.58; P = 0.002; multivariate HR = 1.78; 95% CI 1.05–3.02; P = 0.003). Subgroup analysis revealed that Fusobacterium positive or high abundance in cancer tissues was associated with poor OS in multivariate analysis but had no statistical differences in PFS or DFS in univariate and multivariate analysis. Additionally, P. gingivalis positive in cancer tissue was also associated with worse OS. Conclusions Our meta-analysis suggests that the composition of the oral microbiome may play a significant role in predicting survival outcomes for cancer patients.

To propose a novel model-free data-driven approach based on the voxel-wise mapping of DCE-MRI time-intensity-curve (TIC) profiles for quantifying and visualizing hemodynamic heterogeneity and to validate its potential clinical applications. From December 2018 to July 2022, 259 patients with 325 pathologically confirmed breast lesions who underwent breast DCE-MRI were retrospectively enrolled. Based on the manually segmented breast lesions, the TIC of each voxel within the 3D whole lesion was classified into 19 subtypes based on wash-in rate (nonenhanced, slow, medium, and fast), wash-out enhancement (persistent, plateau, and decline), and wash-out stability (steady and unsteady), and the composition ratio of these 19 subtypes for each lesion was calculated as a new feature set (type-19). The three-type TIC classification, semiquantitative parameters, and type-19 features were used to build machine learning models for identifying lesion malignancy and classifying histologic grades, proliferation status, and molecular subtypes. The type-19 feature-based model significantly outperformed models based on the three-type TIC method and semiquantitative parameters both in distinguishing lesion malignancy (respectively; AUC = 0.875 vs. 0.831, p = 0.01 and 0.875vs. 0.804, p = 0.03), predicting tumor proliferation status (AUC = 0.890 vs. 0.548, p = 0.006 and 0.890 vs. 0.596, p = 0.020), but not in predicting histologic grades (p = 0.820 and 0.970). In addition to conventional methods, the proposed computational approach provides a novel, model-free, data-driven approach to quantify and visualize hemodynamic heterogeneity. Voxel-wise intra-lesion mapping of TIC profiles allows for visualization of hemodynamic heterogeneity and its composition ratio for differentiation of malignant and benign breast lesions. • Voxel-wise TIC profiles were mapped, and their composition ratio was compared between various breast lesions. • The model based on the composition ratio of voxel-wise TIC profiles significantly outperformed the three-type TIC classification model and the semiquantitative parameters model in lesion malignancy differentiation and tumor proliferation status prediction in breast lesions. • This novel, data-driven approach allows the intuitive visualization and quantification of the hemodynamic heterogeneity of breast lesions.

Background:Preoperative non-invasive histologic grading of breast cancer is essential. This study aimed to explore the effectiveness of a machine learning classification method based on Dempster-Shafer (D-S) evidence theory for the histologic grading of breast cancer.Methods:A total of 489 contrast-enhanced magnetic resonance imaging (MRI) slices with breast cancer lesions (including 171 grade Ⅰ, 140 grade Ⅱ, and 178 grade Ⅲ lesions) were used for analysis. All the lesions were segmented by two radiologists in consensus. For each slice, the quantitative pharmacokinetic parameters based on a modified Tofts model and the textural features of the segmented lesion on the image were extracted. Principal component analysis was then used to reduce feature dimensionality and obtain new features from the pharmacokinetic parameters and texture features. The basic confidence assignments of different classifiers were combined using D-S evidence theory based on the accuracy of three classifiers: support vector machine (SVM), Random Forest, and k-nearest neighbor (KNN). The performance of the machine learning techniques was evaluated in terms of accuracy, sensitivity, specificity, and the area under the curve.Results:The three classifiers showed varying accuracy across different categories. The accuracy of using D-S evidence theory in combination with multiple classifiers reached 92.86%, which was higher than that of using SVM (82.76%), Random Forest (78.85%), or KNN (87.82%) individually. The average area under the curve of using the D-S evidence theory combined with multiple classifiers reached 0.896, which was larger than that of using SVM (0.829), Random Forest (0.727), or KNN (0.835) individually.Conclusions:Multiple classifiers can be effectively combined based on D-S evidence theory to improve the prediction of histologic grade in breast cancer.

Background:Quantitative muscle and fat data obtained through body composition analysis are expected to be a new stable biomarker for the early and accurate prediction of treatment-related toxicity, treatment response, and prognosis in patients with lung cancer. The use of these biomarkers can enable the adjustment of individualized treatment regimens in a timely manner, which is critical to further improving patient prognosis and quality of life. We aimed to develop a deep learning model based on attention for fully automated segmentation of the abdomen from computed tomography (CT) to quantify body composition.Methods:A fully automatic segmentation deep learning model was designed based on the attention mechanism and using U-Net as the framework. Subcutaneous fat, skeletal muscle, and visceral fat were manually segmented by two experts to serve as ground truth labels. The performance of the model was evaluated using Dice similarity coefficients (DSCs) and Hausdorff distance at 95th percentile (HD95).Results:The mean DSC for subcutaneous fat and skeletal muscle were high for both the enhanced CT test set (0.93±0.06 and 0.96±0.02, respectively) and the plain CT test set (0.90±0.09 and 0.95±0.01, respectively). Nevertheless, the model did not perform well in the segmentation performance of visceral fat, especially for the enhanced CT test set. The mean DSC for the enhanced CT test set was 0.87±0.11, while the mean DSC for the plain CT test set was 0.92±0.03. We discuss the reasons for this result.Conclusions:This work demonstrates a method for the automatic outlining of subcutaneous fat, skeletal muscle, and visceral fat areas at L3.

BackgroundAmide proton transfer (APT) imaging has been increasingly applied in tumor characterization. However, its value in evaluating breast cancer remains undetermined.PurposeTo assess the diagnostic performance of APT imaging in breast cancer and its association with prognostic histopathologic characteristics.Study TypeProspective.SubjectsEighty‐four patients with breast lesions.Field Strength/SequenceA 3.0 T/single‐shot fast spin echo APT imaging.AssessmentAPTw signal in breast lesion was quantified. Lesion malignancy, T stage, grades, Ki‐67 index, molecular biomarkers (estrogen receptor [ER] expression, progesterone receptor [PR] expression, human epidermal growth factor receptor [HER‐2] expression), molecular subtypes (luminal A, luminal B, triple negative, and HER‐2 enriched) were determined.Statistical TestsStudent t‐test, one‐way analysis of variance, receiver operating characteristic analysis, and Pearson's correlation with P < 0.05 as statistical significance.ResultsAPTw signal was significantly higher in malignant lesions (1.55% ± 1.24%) than in benign lesions (0.54% ± 1.13%), and in grade III lesions than in grade II lesions (1.65% ± 0.84% vs. 0.96% ± 0.96%), and in T2‐ (1.57% ± 0.64%) and T3‐stage lesions (1.54% ± 0.63%) than in T1‐stage lesions (0.81% ± 0.64%) for invasive breast carcinoma of no special type. APTw signal significantly correlated with Ki‐67 index (r = 0.364) but showed no significant difference in groups of ER (P = 0.069), PR (P = 0.069), HER‐2 (P = 0.961), and among molecular subtypes (P = 0.073).Data ConclusionAPT imaging shows potential in differentiating breast lesion malignancy and associates with prognosis‐related tumor grade, T stage, and proliferative activity.Evidence Level2Technical EfficacyStage 2

BackgroundIt remains not well known whether skeletal muscle mass (SMM) loss has any impact on the effectiveness of immune checkpoint inhibitors (ICIs) in patients with advanced lung cancer. We aimed to evaluate the association between SMM and clinical outcome of patients with advanced lung cancer receiving ICIs as first line or second line.Materials and MethodsFrom March 1st, 2019 to March 31st, 2021 at our hospital, 34 patients with advanced lung cancer treated with first-line or second-line ICIs were enrolled retrospectively. The estimation of skeletal muscle index (SMI) for sarcopenia was assessed at the level of the third lumbar vertebra (L3) on computed tomography (CT) images obtained within 4 weeks before initiation of ICIs treatment. The impact of sarcopenia (low SMI) on progression free survival (PFS) was analyzed using Kaplan-Meier method and log-rank tests. The effect of various variables on PFS was evaluated using Cox proportional hazards regression model with univariate and multivariate analysis. The impact on treatment response including objective response rate (ORR) and disease control rate (DCR) and immunotherapy related adverse events (irAEs) between patients with and without sarcopenia was compared by the chi-squared test. The comparison of SMI value between patients with objective response (OR), disease control (DC) and those without OR and DC was used student t-test or Mann-Whitney U test.ResultsBoth in univariate and multivariate analysis, sarcopenia and treatment lines were the predictive factors for PFS (p < 0.05). Patients with sarcopenia had significantly shorter PFS than that of non-sarcopenic ones [6.57 vs. 16.2 months, hazard ratios (HR) = 2.947 and 3.542, and 95% confidence interval (CI): 1.123–13.183 and 1.11–11.308, p = 0.022 and 0.033]. No significant difference in ORR and irAEs was found. Patients with sarcopenia had lower DCR than those without sarcopenia. The mean SMI value of DCR group and non-DCR group was 32.94 ± 5.49 and 44.77 ± 9.06 cm2/m2, respectively (p = 0.008).ConclusionSarcopenia before immunotherapy might be a significant predictor for poor prognosis including shorter PFS and lower DCR in patients with advanced lung cancer treated with ICIs as first line or second line.

Background: To reduce radiation exposure and improve diagnosis in low-dose computed tomography, sev-eral deep learning (DL)-based image denoising methods have been proposed to suppress noise and arti-facts over the past few years. However, most of them seek an objective data distribution approximating the gold standard and neglect structural semantic preservation. Moreover, the numerical response in CT images presents substantial regional anatomical differences among tissues in terms of X-ray absorbency. Methods: In this paper, we introduce structural semantic information for low-dose CT imaging. First, the regional segmentation prior to low-dose CT can guide the denoising process. Second the structural se-mantical results can be considered as evaluation metrics on the estimated normal-dose CT images. Then, a semantic feature transform is engaged to combine the semantic and image features on a semantic fu-sion module. In addition, the structural semantic loss function is introduced to measure the segmentation difference. Results: Experiments are conducted on clinical abdomen data obtained from a clinical hospital, and the semantic labels consist of subcutaneous fat, muscle and visceral fat associated with body physical evaluation. Compared with other DL-based methods, the proposed method achieves better performance on quantitative metrics and better semantic evaluation. Conclusions: The quantitative experimental re-sults demonstrate the promising performance of the proposed methods in noise reduction and structural semantic preservation. While, the proposed method may suffer from several limitations on abnormalities, unknown noise and different manufacturers. In the future, the proposed method will be further explored, and wider applications in PET/CT and PET/MR will be sought. (c) 2022 Elsevier B.V. All rights reserved.