Objective To evaluate the image quality of abdominal dual-energy CT virtual monoenergetic imaging (VMI) in patients with early gastric cancer using titanium alloy clips and assess its effectiveness on reducing metal artifacts. Methods A retrospective study was conducted, including 31 patients with gastric cancer who underwent abdominal dual-energy CT scans with titanium clips inserted in the gastric cavity. Each scan was reconstructed into mixed images (simulated 120 kVp CT) and VMIs with energy levels ranging from 40 keV to 140 keV. Metal artifacts were quantitatively evaluated by measuring the noise values in the lesion and perigastric regions. The contrast-noise ratio (CNR) of the lesion and the corresponding liver tissue was calculated to assess the image quality. Two radiologists independently evaluated the images, considering overall quality, artifact severity, lesion conspicuity, perigastric clarity, and vascular contrast. Results Quantitative analysis revealed that metal artifacts in both the lesion and perigastric regions decreased as the energy level increased. VMIs at 80-140 keV (lesion site) and 90-140 keV (perigastric space) showed significantly fewer artifacts compared to mixed images (P＜0.05). The CNR of lesions remained stable across VMIs at 50-140 keV, while the CNR of normal liver tissue decreased significantly with increasing energy (P＜0.05). In the subjective assessment, VMIs at 80-140 keV had higher artifact scores than mixed images (P＜0.05). VMIs at 70-90 keV provided better lesion conspicuity and perigastric clarity, although vascular contrast decreased significantly with increasing energy (P＜0.05). VMIs at 70-90 keV showed better overall quality (P＜0.05), though not significantly different from mixed images. Conclusions VMIs at 80 keV and 90 keV improve the visibility of lesions and perigastric regions affected by metallic clips, which combined with mixed images can enhance radiologists’ diagnostic accuracy.

Objective:To evaluate the value of radiomics based on contrast-enhanced spectral mammography (CESM) of internal and peripheral regions combined with clinical factors in predicting benign and malignant breast lesions of breast imaging reporting and data system category 4 (BI-RADS 4).Methods:A retrospective analysis was performed on the clinical and imaging data of patients with breast lesions who were treated in Yantai Yuhuangding Hospital (Center 1) Affiliated to Qingdao University from July 2017 to July 2020 and in Fudan University Cancer Hospital (Center 2) from June 2019 to July 2020. Center 1 included 835 patients, all female, aged 17-80 (49±12) years, divided into training set (667 cases) and test set (168 cases) according to the "train-test-split" function in Python software at a ratio of 8∶2; and 49 patients were included from Center 2 as external validation set, all female, aged 34-70 (51±8) years. The radiomics features were extracted from the intralesional region (ITR), the perilesional regions of 5, 10 mm (PTR 5 mm, PTR10 mm) and the intra-and perilesional regions of 5, 10 mm (IPTR 5 mm, IPTR 10 mm) and were selected by variance filtering, SelectKBest algorithm, and least absolute shrinkage and selection operator. Then five radiomics signatures were constructed including ITR signature, PTR 5 mm signature, PTR 10 mm signature, IPTR 5 mm signature, IPTR 10 mm signature. In the training set, univariable and multivariable logistic regressions were used to construct nomograms by selecting radiomics signatures and clinical factors with significant difference between benign and malignant BI-RADS type 4 breast lesions. The efficacy of nomogram in predicting benign and malignant BI-RADS 4 breast lesions was evaluated by the receiver operating characteristic curve and area under the curve (AUC). Decision curve and calibration curve were used to evaluate the net benefit and calibration capability of the nomogram.Results:The nomogram included ITR signature, PTR 5 mm signature, PTR 10 mm signature, IPTR 5 mm signature, age, and BI-RADS category 4 subclassification for differentiating malignant and benign BI-RADS category 4 breast lesions and obtained AUCs of 0.94, 0.92, and 0.95 in the training set, test set, and external validation set, respectively. The calibration curve showed good agreement between the predicted probabilities and actual results and the decision curve indicated a good net benefit of the nomogram for predicting malignant BI-RADS 4 lesions in the training set, test set, and external validation set.Conclusion:The nomogram constructed from the radiomics features of the internal and surrounding regions of CESM breast lesions combined with clinical factors is attributed to differentiate benign from malignant BI-RADS category 4 breast lesions.

Objective:To investigate the differential diagnosis of MRI between male malignant and benign breast lesions.Methods:Totally 34 patients with male breast lesions who underwent breast MRI examination from January 2011 to March 2019 were collected from Shanghai Cancer Center.All images were evaluated by two radiologists who were blinded to pathological results. When there was a disagreement, another independent senior radiologist assessed the imaging features. The imaging features including lesion location, T 1WI signal, T 2WI signal, lesion type and accompanying signs were evaluated. All lesions were confirmed by biopsy or surgical pathology. Twelve patients were in benign group, 22 patients in malignant group. The imaging findings of MRI were recorded and statistically analyzed by univariate analysis (continuous variables were tested by Mann-Whitney U test and categorical variables were tested by Fisher′s exact test). Results:Among the 34 patients, 31 cases clinically touched the mass and 3 cases showed simple nipple bleeding. In MRI signs, breast cancer showed mass-like enhancement (22/22), benign lesions showed non-mass enhancement (7/12), the difference was statistically significant ( P<0.05). And ipsilateral axillary enlarged lymph nodes only appeared in breast cancer, which was significantly different from that in benign lesions ( P<0.05). There was no significant difference in age, lesion location, T 1WI signal, T 2WI signal, skin thickening and nipple invagination between benign and malignant lesions. There was no significant difference in the size, shape and edge of the mass between benign and malignant lesions on MRI ( P>0.05). Conclusions:MRI can distinguish male malignant and benign breast lesions. Most of non-mass enhancement are benign lesion and enlarged lymph nodes are helpful to detect breast cancer, nipple retraction and skin thickening in the diagnosis of male breast cancer are limited.

Objective:To explore the value of different radiomics models based on machine learning in predicting the risk of distant recurrence and metastasis of triple-negative breast cancer after neoadjuvant therapy.Methods:The clinical and imaging data of 150 patients with triple-negative breast cancer (TNBC) confirmed by histopathology were retrospectively analyzed. All patients underwent neoadjuvant chemotherapy and surgical resection from August 2011 to May 2017 in Fudan University Shanghai Cancer Center and Ruijin Hospital, Shanghai Jiao Tong University School of Medicine. One hundred and nine patients from Shanghai Fudan University Shanghai Cancer Center were used as the training group, and 41 patients from Ruijin Hospital, Shanghai Jiao Tong University School of Medicine were used as the validation group. The features were extracted from dynamic contrast-enhanced MRI (DCE-MRI) before treatment and were added with time domain features innovatively. Least absolute shrinkage and selection operator cross validation and recursive feature elimination were applied to select features. Six different supervised machine learning algorithms (logistic regression, linear discriminant analysis, k-nearest neighbor, naive bayesian, decision tree, support vector machine) were used to predict the prognosis. ROC curve, accuracy and F1 measure were used to evaluate the performance of the six algorithms, and also verified by the validation group.Results:The support vector machine algorithm had the best predictive effect in the recurrence and metastasis model based on 15 features, with the highest area under curve (training group was 0.917, validation group was 0.859), and the highest accuracy rate (training group was 87.5%, validation group was 82.9%) and the highest F1 measure (training group was 0.800, validation group was 0.741). In addition, of the 15 imaging features, 12 were the time domain features and 3 were spatial features.Conclusion:With the help of the time domain features and machine learning algorithms, radiomics signatures based on preoperative DCE-MRI can help predict the distant prognosis for TNBC after neoadjuvant chemotherapy and provide support for clinical decision making and follow-up management.

Objective:To evaluate the application value of the histogram features of quantitative parameters from synthetic MRI in predicting the expression of human epidermal growth factor receptor 2 (HER2) in breast invasive ductal carcinoma (IDC) and to compare the prediction efficiency with that of ADC histogram parameters.Methods:A total of 195 patients with breast lesions were prospectively enrolled in the Fudan University Cancer Hospital, from January 2020 to September 2020. All patients underwent preoperative synthetic MRI, DWI and dynamic contrast-enhanced MRI (DCE-MRI). All surgical specimens were confirmed by pathology. The histogram features of the quantitative parameters [T 1, T 2, and proton density (PD)] and ADC values were extracted by PyRadiomics software. Student t test or Mann-Whitney U test were used to compare the histogram characteristics of quantitative parameters (T 1, T 2, and PD) and ADC values between HER2-positive and HER2-negative breast cancers. The diagnostic efficacy of the variables in predicting HER2 expression state was evaluated using the area under curve (AUC) value of ROC. Results:A total of 122 patients with breast IDC were included into analysis, with 31 of HER2-positive and 91 of HER2-negative. There was no significant difference in the clinicopathological characteristics between HER2-positive and HER2-negative breast IDC patients. Univariate analysis showed that there was statistically significant difference in PD-median [79.80 (75.90, 83.90)ms vs. 76.56 (72.59, 79.09) ms, Z=-3.46, P<0.01], PD-mean [78.89 (74.80, 84.01) ms vs. 75.99 (71.70, 78.63) ms, Z=-2.61, P=0.01], PD-Kurtosis [6.45(3.45, 7.54) vs. 5.04 (3.55, 5.58), Z=-2.21, P=0.03], T 1-10 th percentile [731.52 (668.50, 975.39) ms vs. 726.51 (588.38, 852.19) ms, Z=-2.54, P=0.01], T 1-mean [1 161.97 (1 063.56, 1 253.78) ms vs. 1 072.75 (989.39, 1 154.04)ms, Z=-2.21, P=0.03] and ADC-Kurtosis [4.75 (2.72, 5.91) vs. 3.82 (2.69, 4.39), Z=-2.43, P=0.02] between HER2 positive and negative breast IDC patients. Multivariate analysis showed that PD-median ( P=0.004) and T 1-mean ( P=0.004) were independent risk factors for HER2 expression. The ROC curve of HER2 expression predicted by this model showed an AUC was 0.853(95%CI 0.779-0.926), with a sensitivity of 71% and a specificity of 81%. The ROC curve of ADC-Kurtosis for predicting the expression of HER2 showed that the AUC was 0.714 (95%CI 0.611-0.817), with the sensitivity of 45%, and the specificity of 85%. DeLong test showed that the diagnostic efficacy of quantitative parameters from synthetic MRI in predicting the status of HER2 was higher than that of ADC histogram parameters ( Z=2.18, P=0.04). Conclusion:Histogram features of synthetic MRI quantitative parameters contribute to the prediction of HER2 expression status in IDC and may therefore contribute to the determination of individualized anti-HER2 targeted therapy strategies.

Objective:To combine automatic image segmentation technology and machine learning methods to accurately classify and recognize mammography images.Methods:Taking mammography images with clustered pleomorphic calcification as the research object, which were in BI-RADS4 class from the Digital Mammogram Database (DDSM). The region of interest (ROI) of the images was automatically segmented. The characteristic features extracted by wavelet transform, Gabor filter and gray level co-occurrence matrix method were fused. The fused feature parameters were screened based on sensitivity analysis. Using ensemble learning method, the polynomial kernel SVM, random forest and logistic regression classifiers were integrated to form a classifier for automatic classification of mammography images. The ensemble learning method was soft voting integration.Results:The proposed ensemble classifier can efficiently recognize and classify mammography images, and its classification sensitivity, specificity and accuracy on the training set were 99.1%, 99.6% and 99.3%, respectively.Conclusions:The proposed mammography image processing, classification and recognition method can provide assistant detection basis for doctors' clinical judgment, and provide a technical basis for subdividing BI-RADS4 class images.

Objective:To evaluate the value of mammography and MRI combined with clinical features in predicting upgrade to malignancy in high-risk breast lesions.Methods:Data from 230 patients who were diagnosed with high-risk breast lesions and underwent both mammography and MRI examinations before biopsy were analyzed retrospectively from Jan 2017 to Mar 2018 in Fudan University Shanghai Cancer Hospital. The imaging features of both mammography and MRI were analyzed, and the association between mammography, MRI and clinical features were evaluated using pathology as the gold standard. Independent t test and χ 2 test were used to compare the difference of clinical and imaging features between upgrade and non-upgrade groups, using receiver operating characteristic (ROC) curve to test the diagnostic value between mammography and MRI. Binary logistic regression was used to evaluate the correlation between upgrade and clinical, imaging findings. Results:Two hundred and thirty patients had 230 lesions, and 47 cases had atypia upgrade to malignancy during second surgery (upgrade rate was 20.4%). There were statistically significant differences in age, maximum diameter of lesion, and menopausal status between the upgraded and non-upgraded groups ( P<0.05). There was no statistically significant difference in mammographic features between two groups ( P>0.05), while there was statistically significant difference in breast MRI features and background parenchymal enhancement ( P<0.05). For the diagnostic value in predicting upgrade of high-risk lesions, MRI was better than mammography (the areas under ROC curve were 0.913 and 0.606, Z=6.919, P<0.01). Single factor analysis showed that age, lesion size, menopausal status, MRI negative and background parenchymal enhancement on MRI were significantly different for upgrade to malignancy ( P<0.05). Multiple factors analysis showed age and background parenchymal enhancement on MRI were independent factors for predicting upgrade ( P<0.01). Conclusion:For the upgrade to malignancy in high-risk lesions, the diagnostic value of MRI is better than mammography. The elder age and moderate or marked background parenchymal enhancement on MRI may serve as useful predictors of upgrade.

Objective:To evaluate the feasibility of CT radiomics method in predicting outcomes of simultaneous pulmonary nodules in breast cancer patients after treatment.Methods:Patients with breast cancer confirmed by pathology and with simultaneous pulmonary nodules (diameter>5 mm, number≤5) detected by preoperative CT were retrospectively enrolled in this study. Eighty female patients were included (median age: 52, quartile range: 45, 61). The pulmonary nodules (median size: 6.0 mm, quartile range: 5.5, 7.2 mm) were classified into stable group (without change over 2 years) and change group according to follow-up CT findings. The change group was further divided into improved group and progressive group. Eventually, 54 cases were in the stable group, 26 cases were in the change group. One hundred and five texture features were extracted using the python-based pyradiomics package based on preoperative CT images. Stepwise regression was used to exclude features without significant difference in predicting changes of pulmonary nodules. Classifiers model and 5 fold cross validation method were used to obtain the highest performance in predicting outcomes of pulmonary nodules. Receiver operating characteristic (ROC) curve was performed to evaluate the diagnostic performance of the model.Results:After features exclusion and selection, three radiomics features were used to establish classifiers between stable group and change group. It was showed that the linear discriminate analysis was the optimal model with the specificity, sensitivity, accuracy and area under the ROC curve (AUC) as 0.980, 0.460, 0.813 and 0.770 respectively. One radiomics feature was chosen to establish classifiers between improved group and progressive group. The coarse gaussian support vector machine (CGSVM) was the optimal model, with the specificity, sensitivity, accuracy and AUC as 0.540, 0.920, 0.713 and 0.880 respectively.Conclusions:CT radiomics analysis has the potential to predict the outcomes of simultaneous indeterminate pulmonary nodules in breast cancer patients after treatment, and it may contribute to preoperative treatment and postoperative follow-up planning.

OBJECTIVE: To investigate the value of diffusion-weighted magnetic resonance imaging (DW-MRI) as a noninvasive tool to assess salivary gland function for follow-up of patients with radiation-induced xerostomia. MATERIALS AND METHODS: This study included 23 patients with nasopharyngeal carcinoma who had been treated with parotid-sparing radiotherapy (RT). Salivary function was assessed by DW-MRI pre-treatment and one week and one year post-RT, respectively. The maximum apparent diffusion coefficient (ADC) of parotid glands (pADCmax) and the time to peak ADC of parotid glands (pTmax) during stimulation were obtained. Multivariate analysis was used to analyze factors correlated with the severity of radiation-induced xerostomia. RESULTS: The ADCs of parotid and submandibular glands (1.26 ± 0.10 × 10−3 mm2/s and 1.32 ± 0.07 × 10−3 mm2/s pre-RT, respectively) both showed an increase in all patients at one week post-RT (1.75 ± 0.16 × 10−3 mm2/s, p < 0.001 and 1.70 ± 0.16 × 10−3 mm2/s, p < 0.001, respectively), followed by a decrease in parotid glands at one year post-RT(1.57 ± 0.15 × 10−3 mm2/s, p < 0.001) but not in submandibular glands (1.69 ± 0.18 × 10−3 mm2/s, p = 0.581). An improvement in xerostomia was found in 13 patients at one year post-RT. Multivariate analysis revealed 4 significant predictors for the improvement of xerostomia, including dose to parotid glands (p = 0.009, odds ratio [OR] = 0.639), the ADC of submandibular glands (p = 0.013, OR = 3.295), pADCmax (p = 0.024, OR = 0.474), and pTmax (p = 0.017, OR = 0.729) at one week post-RT. CONCLUSION: The ADC value is a sensitive indicator for salivary gland dysfunction. DW-MRI is potentially useful for noninvasively predicting the severity of radiation-induced xerostomia.
Diffusion ; Follow-Up Studies* ; Head and Neck Neoplasms ; Humans ; Magnetic Resonance Imaging* ; Multivariate Analysis ; Odds Ratio ; Parotid Gland ; Radiotherapy ; Salivary Glands* ; Submandibular Gland ; Xerostomia*

Objective To retrospectively investigate the characteristics of background parenchymal enhancement(BPE)in the contralateral breast following neoadjuvant chemotherapy(NAC)and whether BPE could help predict tumor response in early stage of advanced breast cancer. Methods Data from 161 patients who were diagnosed with unilateral breast cancer and then underwent NAC before surgery were analyzed retrospectively from August 2014 to December 2016.All the patients underwent both bilateral breast MRI scan with contrast enhancement. Two experienced radiologists independently categorized the patients' levels of BPE into four categories (1=minimal, 2=mild, 3=moderate, 4=marked) at baseline and after the 2nd cycle of NAC. All the patients were divided in to pathologic complete response (pCR) group and non-pCR group according to the histopathologic tumour response.The status of estrogen receptor(ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) were recorded before underwent NAC.The t test and Pearson chi-squared test were used to compare the baseline characteristics of the pCR and Non-pCR groups.The kappa test was used to test the inter-observer agreement.The Wilcoxon test was used to test for changes of BPE categories after NAC.The Mann-Whitney U test was used to test the difference of BPE between pre-and post-menopausal status. Associations were evaluated using Binary logistic regression models. Results Fifty nine patients achieved pCR, and 102 patients had residual disease (non-pCR). Age, tumor size, distribution of size, menopausal status and lymph node showed no significance between pCR and non-pCR groups(all P>0.05),while only ER/PR status and HER2 status had a significant difference (P>0.05 in both). Inter-observer agreement regarding BPE categorization was moderate and substantial before and after NAC(Kappa value 0.644 and 0.708).The level of BPE was higher in premenopausal than post-menopausal women both at baseline and after the 2nd cycle of NAC(P<0.01). Decreased BPE was observed in 106 cases(premenopausal 60 cases and postmenopausal 46 cases),and no change in BPE were observed in 55 cases (premenopausal 27 cases and postmenopausal 28 cases). A significant reduction in BPE was observed after the 2nd NAC cycle in the overall cases, pre-and post-menopausal cases (all P<0.01). Logistic model showed that hormonal receptor (HR) negative and HER-2 receptor at baseline and the change of BPE after NAC were the independent factors for predicting pCR. Conclusions Regardless of the menopausal status, BPE showed a reduction after NAC, and it can serve as an additional imaging biomarker of tumour response at an early stage of NAC.