西门子医疗系统有限公司

Objective: To evaluate the value of machine learning (ML) models based on biparametric magnetic resonance imaging (bpMRI) for diagnosis of prostate cancer (PCa) and clinically significant prostate cancer (csPCa). Methods: A total of 1 368 patients, aged from 30 to 92 (69.4±8.2) years, from 3 tertiary medical centers in Jiangsu Province were retrospectively collected from May 2015 to December 2020, including 412 cases of csPCa, 242 cases of clinically insignificant prostate cancer (ciPCa) and 714 cases of benign prostate lesions. The data of center 1 and center 2 were randomly divided into training cohort and internal testing cohort at a ratio of 7∶3 by random number sampling without replacement using Python Random package, and the data of center 3 were used as the independent external testing cohort. The training cohort includs 243 cases of csPCa, 135 cases of ciPCa and 384 cases of benign lesions, the internal testing cohort includs 104 cases of csPCa, 58 cases of ciPCa and 165 cases of benign lesions, and the external testing cohort includs 65 cases of csPCa, 49 cases of ciPCa and 165 cases of benign lesions. The radiomics features were extracted on T2-weighted imaging, diffusion-weighted imaging and apparent diffusion coefficient map, and optimal radiomics features were selected by using Pearson correlation coefficient method and analysis of variance. The ML models were built using two ML algorithms, including support vector machine and random forest (RF) and were further tested in the internal testing cohort and external testing cohort. Finally, the PI-RADS scores evaluated by the radiologists were adjusted by the ML models which had superior diagnostic performance, namely adjusted PI-RADS. The receiver operating characteristic (ROC) curves were used to evaluate the diagnostic performance of the ML models and PI-RADS. DeLong test was used to compare the areas under curve (AUC) of models with those of PI-RADS. Results: For PCa diagnosis, in internal testing cohort, the AUC of ML model using RF algorithm and PI-RADS were 0.869 (95%CI: 0.830-0.908) and 0.874 (95%CI: 0.836-0.913), respectively, and the difference between the model and PI-RADS did not reach to the statistical significance (P=0.793). In the external testing cohort, the AUC of model and PI-RADS were 0.845 (95%CI: 0.794-0.897) and 0.915 (95%CI: 0.880-0.951), respectively, and the difference was statistically significant (P=0.01). For csPCa diagnosis, the AUC of ML model using RF algorithm and PI-RADS were 0.874 (95%CI: 0.834-0.914) and 0.892 (95%CI: 0.857-0.927), respectively, in internal testing cohort, and the difference between the model and PI-RADS was not statistically significant (P=0.341). In the external testing cohort, the AUC of model and PI-RADS were 0.876 (95%CI: 0.831-0.920) and 0.884 (95%CI: 0.841-0.926), respectively, and the difference between the model and PI-RADS was not statistically significant (P=0.704). When PI-RADS assessment was adjusted with the assistance of ML models, the specificities increased from 63.0% to 80.0% in the internal testing cohort and from 92.7% to 93.3% in the external test group in diagnosing PCa. In diagnosing csPCa, the specificities increased from 52.5% to 72.6% in the internal testing cohort and from 75.2% to 79.9% in the external testing cohort. Conclusions: The ML models based on bpMRI showed comparable diagnostic performance to PI-RADS assessed by senior radiologists and achieved good generalization ability in both diagnosing PCa and csPCa. The specificities of the PI-RADS were improved by ML models.

Objective: To explore the characteristics of high-resolution computed tomography (HRCT) in diabetes complicated with coronavirus disease 2019 (COVID-19)-associated pneumonia. Materials and Methods: This study included 584 patients (359 males and 225 females), aged between 60~99 years old (mean, (76±9) years), with positive chest computed tomography (CT) findings and diagnosed with COVID-19 in our hospital from December 14, 2022, to January 10, 2023. Of these, 225 patients were diabetic and 359 were non-diabetic. The features of the chest HRCT from patients with diabetes mellitus complicated with COVID-19 and those without diabetes mellitus complicated with COVID-19 were compared. Moreover, 363 patients in the acute stage of COVID-19 (defined as the time interval between onset and CT examination <7 days) were selected for subgroup analysis, and the HRCT characteristics of COVID-19 between the diabetes group and the non-diabetic group in the acute stage. Results: The location, distribution, morphology, and concomitant signs of pulmonary lesions between the two groups of patients with COVID-19 did not differ significantly. Conversely, statistically significant differences in density (fine mesh, uneven density) and lesion margin (fuzzy lesion margin) were detected. In particular, the grid, uneven, and fuzzy signs on lung imaging were significantly higher in the non-diabetic group than that in the diabetic group. Additionally, 54 patients (24%) in the diabetic group and 127 patients (35.38%) in the non-diabetic group demonstrated fine mesh shadows. There were 181 patients (80.44%) in the diabetic group and 313 patients (87.19%) in the non-diabetic group with uneven density. Furthermore, 205 patients (91.11%) in the diabetic group and 344 patients (95.82%) in the non-diabetic group had blurred edges. There was significantly less pulmonary grid shadowing in the acute subgroup with diabetes (35, 24.65%) than in the acute subgroup without diabetes (82, 37.10%). Conclusion: The features of chest HRCT in patients with diabetes mellitus and COVID-19 are mainly exudation, uniform density, and a clear edge, while the interstitial changes are not obvious compared with patients in the non-diabetic group.

Purpose:To explore the optimal monoenergetic level for the observation of carotid artery in-stent lumen by dual-energy CT(DECT).Methods:Forty patients with 51 stents after carotid artery stenting(CAS)who underwent carotid artery CTA examinations by DECT between January 2018 and December 2022 at Renji Hospital were enrolled in the study.Conventional and virtual monoenergetic images(VMI+)in the range of 40-190 keV with an interval of 10 keV were reconstructed.The signal to noise ratio(SNR),contrast to noise ratio(CNR)and noise value were measured.Subjective evaluation of image quality was performed using a 5-grade method.Spearman correlation analysis was used to explore the correlation between energy levels and SNR or CNR.Kruskal-Wallis H rank-sum test was used for the overall analysis between groups,and Dunn-Bonferroni test was used for pairwise comparison within groups.Friedman test was used to compare the differences in the subjective evaluation of image quality in each group.Results:There were significant differences in SNR and CNR among conventional CT images and 40-190 keV VMI+images(P＜0.001).The SNR and CNR of 40 keV images were superior to those of conventional CT images and 60-190 keV VMI+images(P＜0.05).The energy level of 40-190 keV VMI+was negatively correlated with SNR and CNR(SNR:r=-0.763,P＜0.001;CNR:r=-0.696,P＜0.001).The SNR and CNR of 40-50 keV VMI+images were superior to those of conventional CT images(P＜0.001).There were significant differences in noise value among conventional CT images and 40-190 keV VMI+images(P＜0.001).The noise values of 40 keV images were higher than those of conventional CT images and 50-190 keV VMI+images(P＜0.001).The noise values of 40-50 keV VMI+images were superior to those of conventional CT images(P＜0.05).The overall difference in subjective image quality scores among conventional CT images and 40-190 keV VMI+images was with statistical significance(P＜0.001),of which the subjective image quality of 40 keV VMI+was superior to those of 80-190 keV VMI+images(P＜0.001).There were no significant differences in subjective scores between conventional CT images and 40-70 keV VMI+images(P＞0.05).Conclusion:Monoenergetic images at 40 keV is recommended as the optimal level for the observation of carotid artery in-stent lumen.

Objective To study the application of intravoxel incoherent motion(IVIM)imaging and diffusion kurtosis imaging(DKI)in the grading of glioma.Methods The clinical data of 82 patients with brain glioma diagnosed by surgery and puncture pathology in Xianyang Hospital of Yan'an University from May 2022 to April 2023 were ana-lyzed retrospectively,including 12 cases of grade Ⅰ,26 cases of grade Ⅱ,25 cases of grade Ⅲ,and 19 cases of gradeⅣ.All patients underwent conventional MRI plain scan,IVIM,and DKI before operation.IVIM sequence applied 8 groups of b values:0 s/mm2,50 s/mm2,100 s/mm2,150 s/mm2,200 s/mm2,400 s/mm2,800 s/mm2,and 10 000 s/mm2.DKI adopts 30 directions,and b values were 0 s/mm2,1 000 s/mm2,and 2 000 s/mm2.The original data of IVIM and DKI are processed and analyzed by post-processing software,and combined with plain scan and enhanced images,the true dispersion coeffi-cient(D)diagram,perfusion fraction(f)diagram,and false dispersion coefficient(D*)diagram,as well as the parameters of average dispersion coefficient(MD),anisotropic fraction(FA),average dispersion kurtosis(MK),axial kurtosis(AK),and radial kurtosis(RK)are obtained.Receiver operating characteristic(ROC)curve of IVIM and DKI was drawn,and the area under the curve(AUC)was calculated.Results The D*values of tumor areas in patients with gradeⅠ,Ⅱ,Ⅲ,and Ⅳ gliomas were(2.03±0.52)×10-3 mm2/s,(3.32±0.68)×10-3 mm2/s,(4.63±0.82)×10-3 mm2/s,(7.15±0.61)×10-3 mm2/s,respec-tively;the MK values were 0.71±0.14,0.86±0.19,1.03±0.18,and 1.25±0.23,respectively;the higher the grade of glioma,the higher the level of D*and MK,with statistically significant difference(P<0.05).The AUC of D*value and MK value in differentiating high-grade gliomas were 0.840 and 0.821,respectively,indicating that both D*value and MK value had high diagnostic efficiency in differentiating high-grade gliomas.Conclusion There are significant differences in D*value and MK value of patients with different grades of glioma.The tumor of glioma can be accurately graded by IVIM and DKI scanning and related parameters,which can provide important reference for clinical diagnosis and prognosis evaluation.

Objective:To prospectively guide the change of chemotherapy regimen in mouse 5-fluorouracil (5-FU) resistance subcutaneous xenograft tumor model derived from gastric cancer patients by the early changes of MRI apparent diffusion coefficient (ADC), and to compare the difference of tumor load between ADC guided dressing change group and volume guided dressing change group.Methods:From January to June 2020, thirty patient-derived xenografts mouse models were established using 5-FU resistant gastric cancer cells coming from patients, and were randomly divided into experimental group and control group by AdaBoost algorithm, with 15 mice in each group. On the 26th day after transplantation, all mice began chemotherapy with 5-FU as the first-line chemotherapy drug, and underwent MR examination once every two days, including T 2WI and diffusion weighted imaging (DWI). Volumes of tumors were measured using an open-source software ITK-SNAP and values of ADC were measured on ADC maps. According to the change rate of tumor ADC value in the experimental group and the tumor volume growth rate in the control group, the replacement time of chemotherapy drugs was determined, and 5-FU was replaced by paclitaxel. The end point of the experiment was the day that the mice entered the cachexia state. Independent-sample t test was used to compare the difference of tumor load between the two groups. Results:After 5-FU treatment, the ADC value of the two groups both increased. The ADC value began to decline on the 4th day after chemotherapy, and the experimental group continued chemotherapy with paclitaxel instead of 5-FU at this time point. The tumor volume growth rate of the control group increased significantly on the 6th day after chemotherapy (from 8.6% to 16.1%), and the control group used paclitaxel instead of 5-FU chemotherapy at this time point. The observed end point was on the 18th day after chemotherapy. The tumor load of the experimental group [(1.82±0.09) cm 3] was lower than that of the control group [(2.01±0.09) cm 3], and the difference was statistically significant ( t=2.25, P=0.033). On the 16th day after chemotherapy in the experimental group and the 18th day after chemotherapy in the control group, the time of paclitaxel administration in both groups was 12 days. The tumor load in the experimental group [(1.61±0.12) cm 3] was also lower than that in the control group [(2.01±0.09) cm 3], and the difference was statistically significant ( t=2.03, P=0.040). Conclusions:For the subcutaneous transplantation model of 5-FU resistant gastric cancer mice, according to the early changes of tumor ADC value after chemotherapy, the replacement of chemotherapy drugs can obtain a lower tumor load, suggesting that it is a feasible method to optimize the chemotherapy regimen.

Objective:To investigate the prognosis value of baseline contrast-enhanced CT in predicting progression-free survival (PFS) and overall survival (OS) for clinically diagnosed as metastatic far-advanced gastric cancer patients.Methods:Between January 2019 and May 2020, 85 pathologically confirmed gastric adenocarcinoma patients with peritoneal or hepatic metastasis at Shanghai Ruijin Hospital with complete preoperative clinical, image and follow-up data were enrolled in this retrospective study. Clinical factors included performance status (PS) score, tumor location, and tumor serological indicators. Imaging factors included the longest diameter and maximum cross-sectional area of the tumor, CT value, enhancement uniformity, CT extramural venous invasion (ctEMVI), the largest short diameter of the metastatic lymph nodes, confluent lymph nodes, lymph nodes necrosis, fused bulk lymph nodes, the maximum cross-sectional area and CT value of the liver metastases, peritoneal metastasis score, longest diameter of nodules with peritoneal metastasis. Kaplan-Meier survival curve and log-rank test were used to analyze the prognostic differences between groups. Univariate and multivariate Cox proportional hazards regression models were used to identify independent risk factors for PFS and OS.Results:There were significant differences in the maximum cross-sectional area of the tumor, non-contrast CT value, delayed-phase CT value, and delayed-phase CT ratio value between the high- and low-risk groups in PFS ( P<0.05). There were significant differences between the high- and low-risk groups with the maximum cross-sectional area of the tumor in PFS and OS ( P<0.05). In the univariate analysis, the maximum cross-sectional area of tumor, plain-scan CT value, delayed-phase CT value, delayed-phase CT ratio value and the largest short diameter of metastatic lymph nodes were risk factors for PFS ( P<0.05). PS score, CA724, maximum cross-sectional area of the tumor, maximum cross-sectional area of liver metastases, and peritoneal metastasis score were shown as risk factors for OS ( P<0.05). In the multivariate analysis, the maximum cross-sectional area of the tumor and non-contrast CT value were independent risk factors for PFS (HR=0.41, 2.50, P<0.05, 0.006). PS score, CA724 and peritoneal metastasis score were independent risk factors for OS (HR=46.78, 6.26, 92.92, P=0.026, 0.009, 0.007). Conclusions:Tumor size, CT attenuations, and peritoneal metastasis score on baseline CT can be used as independent risk factors for survival in patients with far-advanced gastric cancer with peritoneal or hepatic metastasis. Baseline CT is potentially useful in prediction of the survival status for patients with metastatic far-advanced gastric cancer.