Objective To explore the diagnostic efficacy of machine learning in predicting perineural invasion(PNI)of invasive breast cancer based on MRI imaging features of breast cancer.Methods The data of 294 patients with invasive breast cancer confirmed by surgical pathology were retrospectively analyzed,and the patients were randomly divided into training set(205 cases,PNI 77 cases)and validation set(89 cases,PNI 33 cases)at a ratio of 7∶3.10 machine learning models were constructed by selecting training set clinical and radiographic features using single factor logistic regression.The area under the curve(AUC),accuracy(ACC),sensitivity(SE),specificity(SP),positive predictive value(PPV),and negative predictive value(NPV)were used to evaluate the predictive effi-cacy of different models for PNI,and the best model was determined.SHapley Additive exPlanation(SHAP)was used to visuaize the diagnosis process of the model.Results In the validation set,the multi-layer perceptron(MLP)model performed best,with AUC,ACC,SE,SP,PPV,and NPV of 0.91,0.89,0.79,0.95,0.90,and 0.88,respectively.Conclusion The model of MRI imaging fea-tures of breast cancer constructed by MLP machine learning model can effectively predict the preoperative PNI of invasive breast cancer.

Objective To explore the risk prediction value of gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid enhanced magnetic resonance imaging(Gd-EOB-MRI)liver imaging reporting and data system version 2018(LI-RADS v2018)with contrast-enhanced ultrasound(CEUS)LI-RADS version 2017(LI-RADS v2017)in high-risk patients with hepatocellular carcinoma(HCC).Methods The clinical and imaging data of 122 high-risk patients for HCC(with a total of 134 liver lesions)who underwent both Gd-EOB-MRI and CEUS examination at the same time and obtained pathological results within one month were retrospectively col-lected.The nodules were classified according to the CT/MRI LI-RADS v2018 and CEUS LI-RADS v2017 criteria,and the LI-RADS classification results of the two imaging methods were subjected to Cohen's Kappa test.Using pathological results as the gold stand-ard,the diagnostic efficacy of LI-RADS v2018 and LI-RADS v2017 with LR-5 as the standard for HCC was calculated separately.Results The overall consistency between the Gd-EOB-MRI LI-RADS v2018 and CEUS LI-RADS v2017 classification standards was good(Kappa=0.691,P＜0.001).Using LR-5 as the standard for diagnosing HCC,the sensitivity of Gd-EOB-MRI and CEUS was 84.7％and 81.2％,the specificity was 79.6％and 73.5％,the positive predictive value was 87.8％and 84.1％,the negative predictive value was 75.0％and 69.2％,and the accuracy was 82.8％and 78.4％,respectively.There was no statistically significant difference in the diagnostic efficacy of the diagnosis of HCC by LR-5 between the two imaging methods(P＞0.05).Conclusion The Gd-EOB-MRI LI-RADS v2018 and CEUS LI-RADS v2017 classifica-tion standards show good overall agreement.The diagnostic efficacy of Gd-EOB-MRI for HCC using LR-5 classification is better than that of CEUS.

Objective To explore the diagnostic efficacy of machine learning in predicting perineural invasion(PNI)of invasive breast cancer based on MRI imaging features of breast cancer.Methods The data of 294 patients with invasive breast cancer confirmed by surgical pathology were retrospectively analyzed,and the patients were randomly divided into training set(205 cases,PNI 77 cases)and validation set(89 cases,PNI 33 cases)at a ratio of 7∶3.10 machine learning models were constructed by selecting training set clinical and radiographic features using single factor logistic regression.The area under the curve(AUC),accuracy(ACC),sensitivity(SE),specificity(SP),positive predictive value(PPV),and negative predictive value(NPV)were used to evaluate the predictive effi-cacy of different models for PNI,and the best model was determined.SHapley Additive exPlanation(SHAP)was used to visuaize the diagnosis process of the model.Results In the validation set,the multi-layer perceptron(MLP)model performed best,with AUC,ACC,SE,SP,PPV,and NPV of 0.91,0.89,0.79,0.95,0.90,and 0.88,respectively.Conclusion The model of MRI imaging fea-tures of breast cancer constructed by MLP machine learning model can effectively predict the preoperative PNI of invasive breast cancer.

Objective To explore the risk prediction value of gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid enhanced magnetic resonance imaging(Gd-EOB-MRI)liver imaging reporting and data system version 2018(LI-RADS v2018)with contrast-enhanced ultrasound(CEUS)LI-RADS version 2017(LI-RADS v2017)in high-risk patients with hepatocellular carcinoma(HCC).Methods The clinical and imaging data of 122 high-risk patients for HCC(with a total of 134 liver lesions)who underwent both Gd-EOB-MRI and CEUS examination at the same time and obtained pathological results within one month were retrospectively col-lected.The nodules were classified according to the CT/MRI LI-RADS v2018 and CEUS LI-RADS v2017 criteria,and the LI-RADS classification results of the two imaging methods were subjected to Cohen's Kappa test.Using pathological results as the gold stand-ard,the diagnostic efficacy of LI-RADS v2018 and LI-RADS v2017 with LR-5 as the standard for HCC was calculated separately.Results The overall consistency between the Gd-EOB-MRI LI-RADS v2018 and CEUS LI-RADS v2017 classification standards was good(Kappa=0.691,P＜0.001).Using LR-5 as the standard for diagnosing HCC,the sensitivity of Gd-EOB-MRI and CEUS was 84.7％and 81.2％,the specificity was 79.6％and 73.5％,the positive predictive value was 87.8％and 84.1％,the negative predictive value was 75.0％and 69.2％,and the accuracy was 82.8％and 78.4％,respectively.There was no statistically significant difference in the diagnostic efficacy of the diagnosis of HCC by LR-5 between the two imaging methods(P＞0.05).Conclusion The Gd-EOB-MRI LI-RADS v2018 and CEUS LI-RADS v2017 classifica-tion standards show good overall agreement.The diagnostic efficacy of Gd-EOB-MRI for HCC using LR-5 classification is better than that of CEUS.

Objective Toinvestigatetheconsistencyofultrasoundand MSCTindiagnosing microvascularinvasion (MVI)in hepatocellularcarcinoma,andtocomparethediagnosticefficiencyofbothtwomethods.Methods TheultrasoundandMSCTdataof 82patientswithhepatocellularcarcinomawerecollected.Accordingtothepostoperativepathologicalresults,theconsistencyanddiagnostic efficiencyofultrasoundandMSCTindiagnosingMVIinhepatocellularcarcinomawerecompared.Results Accordingtothepostoperative pathologicalfindingsof82patients,30caseswerepositiveofMVI,while52caseswerenegative.ThepreoperativeMSCTandultrasound examinations had a strong consistency (Cohen’s Kappa＝0.829 ,P＜0.001 ).The sensitivity ,specificity and accuracy of preoperative ultrasoundinpredictingtheMVIinhepatocellularcarcinomawere76.67% (23/30),67.31% (35/52)and70.73% (58/82),respectively. Thesensitivity,specificityandaccuracyofpreoperativeMSCTinpredictinghepatocellularcarcinomaMVIwere83.33% (25/30),73.08%(38/52)and75.61% (63/82),respectively.Conclusion Preoperativeultrasoundand MSCThavegoodconsistencyandhighdiagnostic efficiencyindiagnosing MVIinhepatocellularcarcinoma.

Objective To investigate the value of molecular probe USPIO-PEG-sLeX in metastasis model of nasopharyngeal carcinoma. Methods According to the pathological results of transplanted tumors of nude mice with nasopharyngeal carcinoma,the tumor-bearing mice were divided into metastatic group and non-metastatic group.T2 ? values of the transplanted tumors at four different time points were measured before and after the injection of USPIO-PEG-sLeX(t0＝plain scan,t1＝scan immediately after enhancement,t2＝1 hour post-enhancement scan,t3＝2 hours post-enhancement scan).ΔT2 ? and enhancing rates at t1 ,t2,t3 were calculated.Results The plain T2 ? values of the metastatic group and the non-metastatic group were 22.25±8.08 and 27.01±9.45,respectively.There was no significant difference between the two groups at t0 (P＞0.05).The T2 ? values at t1,t2 and t3 were 11.57±4.02 and 24.82±7.84,10.09±4.88 and 24.15±8.74,12.46±5.63,23.42±7.12 respectively in the metastatic and non-metastatic groups,differences were statistically significant (P values were 0.005,0.008 and 0.01 9,respectively).ΔT2 ? values at time t1 ,t2 and t3 in nude mice in the metastatic and non-metastatic groups were 10.69±6.23,3.86±2.20,12.1 7±8.67 and 2.87±1.37,9.80±3.03,4.32±2.28 respectively.The differences were statistically significant (P values were 0.027,0.048 and 0.043,respectively).The enhancement rates at t1,t2 and t3 in nude mice in metastatic and non-metastatic groups were 45.98±14.03,7.10±5.18,5 1.1 5±22.70 and 1 1.04±6.01 ,44.05 ±13.92,12.09 ±7.54,respectively,with statistically significant differences (P values were 0.001,0.005 and 0.007,respectively).After two-way repeated-measures analysis of variance,there was a statistically significant difference in T2 ? values between the metastatic and non-metastatic groups (P＝0.02＜0.05).There was a statistically significant difference in T2 ? values at different times (P＝0.001＜0.05).There was an interaction between the metastatic and non-metastatic groups and the scan time (P＝0.043＜0.05).Conclusion USPIO-PEG-sLeX molecular probe can non- invasively monitor E-selectin expression,and has a good application prospect in monitoring the metastasis of nasopharyngeal carcino-ma xenograft models.

Objective To investigate the value of spectral CT imaging in diagnosis of axillary lymphatic metastasis of breast carcinoma. Methods 27 cases of breast carcinoma who met the criteria underwent dual-phase enhanced spectra CT scan.The axillary lymph nodes were matched one-to-one between CT images and postoperative pathology.The four parameters,including NIC-A,NIC-V, λHU-A,and λHU-V were analyzed among primary tumors,metastasis and non-metastasis lymph nodes.The area under receiver operating characteristic (ROC)curves of the four parameters were used to evaluate the diagnostic accuracy in axillary lymphatic metastasis.Results The four parameters showed significant differences among the primary tumors,metastasis lymph nodes and non-metastasis lymph nodes. The four parameters were also significantly different between the primary tumors and non-metastasis lymph nodes,between metastasis lymph nodes and non-metastasis lymph nodes.Only NIC-V was significantly different between the primary tumors and metastasis lymph nodes.When NIC-A value of 0.148 was used as the threshold for diagnosing lymphatic metastasis,the sensitivity,specificity and the area under the curve were 80.00%,84.62% and 0.888,respectively.Conclusion The parameters of spectral CT imaging can significantly improve the diagnostic accuracy of axillary lymphatic metastasis of breast carcinoma,which can be used as a reference for clinical diagnosis.

Objective To analyze the MRI findings of pure mucinous breast carcinoma and to investigate the causes of misdiagnosis.Methods MRI findings of 1 3 patients with pure mucinous breast carcinoma were analyzed retrospectively.All patients were verified by surgery and histopathology.Results All the 13 cases were presented as a mass,among which 12 cases (92.3%)had lobular contour and 1 (7.7%) had irregular shape.11 cases (84.6%)showed low or iso-signal intensity and 2 (15.4%)showed heterogeneous iso or high intensity on T1WI.All the 13 cases showed high signal intensity on fat suppression T2WI,and 5 cases(38.5%)showed low signal fiber separation inside. 5 cases (38.5%)showed homogeneous high signal intensity and 8 cases (61.5%)showed heterogeneous high signal intensity on diffusion weighted imaging(DWI)and apparent diffusion coefficient(ADC)map,the ADC value was (2.002±0.478)× 10-3mm2/s.11 cases (84.6%)showed rim enhancement,among which 8(61.5%)showed persistent enhancement and 2 (15.4%)showed heterogeneous enhancement.Time-signal intensity cure(TIC)of 11 cases(84.6%)were monophasic or biphasic and 2 cases(15.4%0)were wash-out type. 5 cases(38.5%)were misdiagnosed on preoperative MRI.Conclusion There are several MRI features in pure mucinous breast carcinoma,such as lobular contour,high signal intensity on DWI and ADC map,persistent enhancement and monophasic or biphasic on TIC.Comprehending these MRI features is the key to reduce misdiagnosis.

Objective To investigate the correlation between apparent diffusion coefficient (ADC)of diffusion weighted imaging (DWI)and prognostic factors in rectal cancer.Methods 5 5 patients with rectal cancer were confirmed pathologically.Conventional pelvic MRI and DWI examination were performed,and the mean ADC values of tumor were measured preoperatively.The patients were divided into two groups with or without lymph node metastasis,and were also divided into four groups with negative,weakly positive,positive or strongly positive expression of EGFR in rectal cancer.The ADC values were calculated in each group,and the correlation of ADC values with the lymph node status and EGFR expression classification were analyzed.Results In 5 5 patients with colorectal cancer,there were 13 lesions with lymph node metastasis and the positive expression rate of EGFR was 67.2%.There were no significant difference in mean ADC value between the groups with and without lymph node metastasis (P=0.342).The number of lesions with negative,weakly positive,positive and strongly positive EGFR expression were 18,15,12 and 10.The difference in the mean ADC values among negative,weakly positive,positive and strongly positive expression groups of prognostic factor EGFR was not significantly different (P=0.412).There were also no correlations in the prognostic factors mentioned above (r=0.183 and -0.324,all P>0.05).Conclusion The ADC value can not be used to predict the prognosis and to provide more valu-able information for individualized therapy in patients with rectal carcinoma,which needs further studiy in the future.

Objective To explore the value of the molecular probe USPIO-PEG-sLeX on nasopharyngeal carcinoma xenograft in nude mice.Methods The USPIO nanoparticles was synthesized by physical deposition method,and which was modified by PEG to synthesize USPIO-PEG-sLeX .The nude mice of nasopharyngeal carcinoma xenograft were divided into experimental and control groups.USPIO-PEG-sLeX and USPIO-PEG were injected into nude mice of experimental and control groups by caudal vein,respectively.MR T2 mapping imaging was scanned before and after the injection,and analyzed the changes of T2 values between experimental and control groups. Results USPIO-PEG-sLeX had a good representation.The non-enhanced T2 values between control and experimental group had no statistical significance (P >0.05).However,T2 values of the mice in two groups before and after injections were statistically significant (P <0.05);and T2 values of experimental group were much lower than that of the control group after the injection,additionally,the difference of enhanced rate between the two groups was statistically significant (P <0.05).Conclusion USPIO-PEG-sLeX magnetic nanoparticles is potential to be a targeted contrast agent to ELAM-1 expression of nasopharyngeal carcinoma,and can be valuable in non-invasive dynamic monitoring the expression of ELAM-1.