Objective:To explore the effectiveness of a hysteroscopic endometrial lesion diagnosis model de-veloped based on deep learning(DL)algorithm combined with gradient-weighted class activation mapping(Grad-CAM)visualization technology.Methods:303 hysteroscopy videos(4781 images)of 291 patients who un-derwent hysteroscopy examination in the Department of Gynecology,Renmin Hospital of Wuhan University from June 1,2021 to December 31,2022 were selected.The dataset was divided into a training set(3703 images)and a test set(1078 images)by weight sampling method.After the training set was used for model learning and train-ing,two model architectures,residual neural network(ResNet18)and efficient neural network(EfficientNet-B0),were selected to verify the model in the test set by five-class and two-class classification tasks,respectively.Tak-ing histopathology as the gold standard,the diagnostic efficacy was evaluated to select the optimal model,and the Grad-CAM layer was embedded in the optimal model to output hysteroscopy images of Grad-CAM.Results:①In the five-class classification tasks,the accuracy of EfficientNet-B0 model(93.23％)was higher than that of Res-Net18 model(84.23％);the area under the curve(AUC)of EfficientNet-B0 model in the diagnosis of five disea-ses,including atypical endometrial hyperplasia,endometrial polyps,endometrial cancer,endometrial atypical hy-perplasia,and submucous myoma,was slightly higher than that of ResNet18 model,and the AUC of both models was almost above 0.980.②In the binary classification task of accuracy and the evaluation of specificity,the two models were similar,both above 93.00％,and the sensitivity of EfficientNet-B0 model(91.14％)was significantly better than that of ResNet18 model(77.22％).③EfficientNet-B0 model combined with Grad-CAM algorithm could identify the abnormal areas in the image.After biopsy and pathological examination,it was confirmed that about 95％of the marked areas in the model's output heatmap were lesion areas.Conclusions:The hysteroscopy di-agnostic model developed by EfficientNet-B0 model combined with Grad-CAM has high diagnostic accuracy,sen-sitivity,and specificity,and has application value in the diagnosis of endometrial lesions.

Objective:To analyze the hepatobiliary phase (HBP) image manifestation classification and pathological features of nodules in nodules accompanied by hepatocellular carcinoma (NIN-HCC).Methods:Twenty-five cases cases (27 lesions) with cirrhosis who were confirmed as NIN-HCC by surgical pathology and underwent gadoxetate disodium-enhanced MRI examination before surgery at Nantong Third Hospital affiliated with Nantong University from July 2015 to November 2022 were retrospectively enrolled. The size, signal intensity, enhancement pattern, and pathological features of internal and external nodules were analyzed in NIN-HCC. The lesions score were recorded according to the 2018 version of the Liver Imaging Reporting and Data Systems (LI-RADS) classification criteria. NIN-HCCs were grouped and typed according to the different HBP signal intensities of the inner and outer nodules. The independent-samples t-test, Mann-Whitney U test or Fisher's exact probability method were used to compare the differences in imaging features and LI-RADS scores between the groups. The Spearman correlation coefficient was used to evaluate the correlation between the pathological differentiation degree of internal and external nodules and the HBP signal intensity. The Kaplan-Meier curve was used to analyze recurrence-free survival (RFS) following NIN-HCC surgery. Results:The internal nodules of the 27 NIN-HCCs showed altered hypervascularity with a maximum diameter of (13.2±5.5) mm during the arterial phase. 51.9% (14/27) and 48.1% (13/27) showed "fast in and fast out" and fast in and slow out"enhancement patterns. The external nodules showed altered hypovascularity with a maximum diameter of (25.7±7.3) mm, and 13 (48.1%) of them were accompanied to manifest during the arterial phase. NIN-HCC was divided into two groups according to the signal intensity of HBP of the outer nodules with the background liver parenchyma signal intensity as a reference: the hyposignal group ( n=17, 63.0%) and the isosignal group ( n=10, 37.0%). The hyposignal group and the isosignal group were divided into A~C type and D~F type, a total of six types, according to the hypo, iso, and hyper signals of the inner nodules and the signal intensity of the outer nodules as a reference. Within the hyposignal group, 7.4% (2/27) of the inner nodules showed hyposignal (type A), 37.0% (10/27) showed isosignal (type B), and 18.5% (5/27) showed hypersignal (type C). Within the isosignal group, 29.6% (8/27) of the inner nodules showed hyposignal (type D), 7.4% (2/27) showed isosignal (type E), and there was no hypersignal (type F). 40.7% (11/27) of the lesions were LR-4 in LI-RADS score, and 59.3% (16/27) were LR-5. There was no statistically significant difference ( P>0.05) in the maximum diameter, enhancement pattern, and LI-RADS score of internal and external nodules between the hypo and iso signal group. Histologically, NIN-HCC showed fine trabecular/pseudoglandular duct type without microvascular invasion, among which the inner nodules were mainly moderately differentiated HCC, and the outer nodules were mainly well-differentiated HCC. The degree of differentiation between the inner and outer nodules and the HBP signal intensity had no statistically significant difference ( r=0.290, P=0.143; r=0.079, P=0.697). The median RFS follow-up time after NIN-HCC radical resection was 31.7 months, and the cumulative RFS rates at 1, 3, and 5 years were 96.0%, 76.0%, and 64.0%, respectively. Conclusions:NIN-HCC can serve as a morphological marker for early-stage diagnosis of multi-step cancer evolution in HCC, with certain imaging and pathological features. HBP imaging classification is helpful to enhance the diagnostic recognition of this disease.

Background and aims:Liver cirrhosis is a complex disease that may result in increased morbidity and mortality following bariatric surgery(BS).This study aimed to explore the outcome of BS in patients with unexpected cirrhosis,focusing on postoperative complications and the progression of liver disease. Methods:A retrospective study of bariatric patients with cirrhosis from four centers in China between 2016 and 2023 was conducted,with follow-up for one year after BS.The primary outcome was the safety of BS in patients with unexpected cirrhosis,while the secondary outcome was the metabolic efficacy of BS in this group postoperatively. Results:A total of 47 patients met the study criteria,including 46 cases of Child-Pugh class A cirrhosis and 1 case of Child-Pugh B.Pathological examination confirmed nodular cirrhosis in 21 patients(44.68％),pseudolobule formation in 1 patient(2.13％),lipedema degeneration with inflammatory cell infiltration in 3 patients(6.38％),and chronic hepatitis in 1 patient(2.13％).The average percentage of total weight loss was 29.73±6.53％at one year postoperatively.During the 30-day postoperative period,the complication rate was 6.38％,which included portal vein thrombosis,gastrointestinal bleeding,and intra-abdominal infection.Moreover,no cases of liver decompensation or mortality were reported during the follow-up period.The remission rates of comorbidities among 41 patients one year after surgery were as fol-lows:dyslipidemia 100％,type 2 diabetes 82.61％,hypertension 84.62％,and obstructive sleep apnea syndrome 85.71％. Conclusions:BS can be safely performed in patients with unexpected cirrhosis in the compensated stage of liver disease,with low postoperative morbidity and no mortality observed during one-year follow-up.

Objective To investigate the clinical effect of anterior cervical discectomy and fusion(ACDF)in the treatment of cervical spondylosis of vertebral artery type(CSA).Methods The clinical data of 42 patients with CSA from January 2020 to January 2022 were retrospectively analyzed.There were 25 males and 17 females,aged from 30 to 74 years old with an average of(53.9±11.0)years old.There were 18 cases with single-segment lesions,17 cases with two-segment lesions,and 7 cases with three-segment lesions.The American Academy of Otolaryngology-Head and Neck Surgery's Hearing and Balance Committee score(CHE),the Neck Disability Index(NDI)and the cervical curvature Cobb angle were recorded before surgery and after surgery at 6 months.Results All 42 ACDF patients were followed up for 6 to 30 months with an average of(14.0±5.2)months.The operative time ranged from 95 to 220 min with an average of(160.38±36.77)min,the intraoperative blood loss ranged from 30 to 85 ml with an average of(53.60±18.98)ml.Tow patients had mild postoperative dysphagia,which improved with symptomatic treatment such as nebulized inhalation.CHE score decreased from(4.05±0.96)preoperatively to(2.40±0.70)at 6 months postoperatively(t=12.97,P＜0.05).The number of improved vertigo at 6 months postoperatively was 38,with an im-provement rate of 90.5％.NDI score was reduced from(34.43±8.04)preoperatively to(20.76±3.91)at 6 months postopera-tively(t=1 1.83,P＜0.05).The cervical curvature Cobb angle improved from(8.04±6.70)° preoperatively to(12.42±5.23)° at 6 months postoperatively(t=-15.96,P＜0.05).Conclusion The ACDF procedure has outstanding clinical efficacy in treating CSA.The operation can rapidly relieve patients'episodic vertigo symptoms by relieving bony compression and reconstructing cervical curvature.However,it is necessary to strictly grasp the indications for surgery and clarify the causes of vertigo in pa-tients,and ACDF surgery is recommended for CSA patients for whom conservative treatment is ineffective.

Objective:To explore the effectiveness of a hysteroscopic endometrial lesion diagnosis model de-veloped based on deep learning(DL)algorithm combined with gradient-weighted class activation mapping(Grad-CAM)visualization technology.Methods:303 hysteroscopy videos(4781 images)of 291 patients who un-derwent hysteroscopy examination in the Department of Gynecology,Renmin Hospital of Wuhan University from June 1,2021 to December 31,2022 were selected.The dataset was divided into a training set(3703 images)and a test set(1078 images)by weight sampling method.After the training set was used for model learning and train-ing,two model architectures,residual neural network(ResNet18)and efficient neural network(EfficientNet-B0),were selected to verify the model in the test set by five-class and two-class classification tasks,respectively.Tak-ing histopathology as the gold standard,the diagnostic efficacy was evaluated to select the optimal model,and the Grad-CAM layer was embedded in the optimal model to output hysteroscopy images of Grad-CAM.Results:①In the five-class classification tasks,the accuracy of EfficientNet-B0 model(93.23％)was higher than that of Res-Net18 model(84.23％);the area under the curve(AUC)of EfficientNet-B0 model in the diagnosis of five disea-ses,including atypical endometrial hyperplasia,endometrial polyps,endometrial cancer,endometrial atypical hy-perplasia,and submucous myoma,was slightly higher than that of ResNet18 model,and the AUC of both models was almost above 0.980.②In the binary classification task of accuracy and the evaluation of specificity,the two models were similar,both above 93.00％,and the sensitivity of EfficientNet-B0 model(91.14％)was significantly better than that of ResNet18 model(77.22％).③EfficientNet-B0 model combined with Grad-CAM algorithm could identify the abnormal areas in the image.After biopsy and pathological examination,it was confirmed that about 95％of the marked areas in the model's output heatmap were lesion areas.Conclusions:The hysteroscopy di-agnostic model developed by EfficientNet-B0 model combined with Grad-CAM has high diagnostic accuracy,sen-sitivity,and specificity,and has application value in the diagnosis of endometrial lesions.

Objective:To explore the effectiveness of a hysteroscopic endometrial lesion diagnosis model de-veloped based on deep learning(DL)algorithm combined with gradient-weighted class activation mapping(Grad-CAM)visualization technology.Methods:303 hysteroscopy videos(4781 images)of 291 patients who un-derwent hysteroscopy examination in the Department of Gynecology,Renmin Hospital of Wuhan University from June 1,2021 to December 31,2022 were selected.The dataset was divided into a training set(3703 images)and a test set(1078 images)by weight sampling method.After the training set was used for model learning and train-ing,two model architectures,residual neural network(ResNet18)and efficient neural network(EfficientNet-B0),were selected to verify the model in the test set by five-class and two-class classification tasks,respectively.Tak-ing histopathology as the gold standard,the diagnostic efficacy was evaluated to select the optimal model,and the Grad-CAM layer was embedded in the optimal model to output hysteroscopy images of Grad-CAM.Results:①In the five-class classification tasks,the accuracy of EfficientNet-B0 model(93.23％)was higher than that of Res-Net18 model(84.23％);the area under the curve(AUC)of EfficientNet-B0 model in the diagnosis of five disea-ses,including atypical endometrial hyperplasia,endometrial polyps,endometrial cancer,endometrial atypical hy-perplasia,and submucous myoma,was slightly higher than that of ResNet18 model,and the AUC of both models was almost above 0.980.②In the binary classification task of accuracy and the evaluation of specificity,the two models were similar,both above 93.00％,and the sensitivity of EfficientNet-B0 model(91.14％)was significantly better than that of ResNet18 model(77.22％).③EfficientNet-B0 model combined with Grad-CAM algorithm could identify the abnormal areas in the image.After biopsy and pathological examination,it was confirmed that about 95％of the marked areas in the model's output heatmap were lesion areas.Conclusions:The hysteroscopy di-agnostic model developed by EfficientNet-B0 model combined with Grad-CAM has high diagnostic accuracy,sen-sitivity,and specificity,and has application value in the diagnosis of endometrial lesions.

Objective:To explore the effectiveness of a hysteroscopic endometrial lesion diagnosis model de-veloped based on deep learning(DL)algorithm combined with gradient-weighted class activation mapping(Grad-CAM)visualization technology.Methods:303 hysteroscopy videos(4781 images)of 291 patients who un-derwent hysteroscopy examination in the Department of Gynecology,Renmin Hospital of Wuhan University from June 1,2021 to December 31,2022 were selected.The dataset was divided into a training set(3703 images)and a test set(1078 images)by weight sampling method.After the training set was used for model learning and train-ing,two model architectures,residual neural network(ResNet18)and efficient neural network(EfficientNet-B0),were selected to verify the model in the test set by five-class and two-class classification tasks,respectively.Tak-ing histopathology as the gold standard,the diagnostic efficacy was evaluated to select the optimal model,and the Grad-CAM layer was embedded in the optimal model to output hysteroscopy images of Grad-CAM.Results:①In the five-class classification tasks,the accuracy of EfficientNet-B0 model(93.23％)was higher than that of Res-Net18 model(84.23％);the area under the curve(AUC)of EfficientNet-B0 model in the diagnosis of five disea-ses,including atypical endometrial hyperplasia,endometrial polyps,endometrial cancer,endometrial atypical hy-perplasia,and submucous myoma,was slightly higher than that of ResNet18 model,and the AUC of both models was almost above 0.980.②In the binary classification task of accuracy and the evaluation of specificity,the two models were similar,both above 93.00％,and the sensitivity of EfficientNet-B0 model(91.14％)was significantly better than that of ResNet18 model(77.22％).③EfficientNet-B0 model combined with Grad-CAM algorithm could identify the abnormal areas in the image.After biopsy and pathological examination,it was confirmed that about 95％of the marked areas in the model's output heatmap were lesion areas.Conclusions:The hysteroscopy di-agnostic model developed by EfficientNet-B0 model combined with Grad-CAM has high diagnostic accuracy,sen-sitivity,and specificity,and has application value in the diagnosis of endometrial lesions.

Objective:To explore the effectiveness of a hysteroscopic endometrial lesion diagnosis model de-veloped based on deep learning(DL)algorithm combined with gradient-weighted class activation mapping(Grad-CAM)visualization technology.Methods:303 hysteroscopy videos(4781 images)of 291 patients who un-derwent hysteroscopy examination in the Department of Gynecology,Renmin Hospital of Wuhan University from June 1,2021 to December 31,2022 were selected.The dataset was divided into a training set(3703 images)and a test set(1078 images)by weight sampling method.After the training set was used for model learning and train-ing,two model architectures,residual neural network(ResNet18)and efficient neural network(EfficientNet-B0),were selected to verify the model in the test set by five-class and two-class classification tasks,respectively.Tak-ing histopathology as the gold standard,the diagnostic efficacy was evaluated to select the optimal model,and the Grad-CAM layer was embedded in the optimal model to output hysteroscopy images of Grad-CAM.Results:①In the five-class classification tasks,the accuracy of EfficientNet-B0 model(93.23％)was higher than that of Res-Net18 model(84.23％);the area under the curve(AUC)of EfficientNet-B0 model in the diagnosis of five disea-ses,including atypical endometrial hyperplasia,endometrial polyps,endometrial cancer,endometrial atypical hy-perplasia,and submucous myoma,was slightly higher than that of ResNet18 model,and the AUC of both models was almost above 0.980.②In the binary classification task of accuracy and the evaluation of specificity,the two models were similar,both above 93.00％,and the sensitivity of EfficientNet-B0 model(91.14％)was significantly better than that of ResNet18 model(77.22％).③EfficientNet-B0 model combined with Grad-CAM algorithm could identify the abnormal areas in the image.After biopsy and pathological examination,it was confirmed that about 95％of the marked areas in the model's output heatmap were lesion areas.Conclusions:The hysteroscopy di-agnostic model developed by EfficientNet-B0 model combined with Grad-CAM has high diagnostic accuracy,sen-sitivity,and specificity,and has application value in the diagnosis of endometrial lesions.

Objective:To explore the effectiveness of a hysteroscopic endometrial lesion diagnosis model de-veloped based on deep learning(DL)algorithm combined with gradient-weighted class activation mapping(Grad-CAM)visualization technology.Methods:303 hysteroscopy videos(4781 images)of 291 patients who un-derwent hysteroscopy examination in the Department of Gynecology,Renmin Hospital of Wuhan University from June 1,2021 to December 31,2022 were selected.The dataset was divided into a training set(3703 images)and a test set(1078 images)by weight sampling method.After the training set was used for model learning and train-ing,two model architectures,residual neural network(ResNet18)and efficient neural network(EfficientNet-B0),were selected to verify the model in the test set by five-class and two-class classification tasks,respectively.Tak-ing histopathology as the gold standard,the diagnostic efficacy was evaluated to select the optimal model,and the Grad-CAM layer was embedded in the optimal model to output hysteroscopy images of Grad-CAM.Results:①In the five-class classification tasks,the accuracy of EfficientNet-B0 model(93.23％)was higher than that of Res-Net18 model(84.23％);the area under the curve(AUC)of EfficientNet-B0 model in the diagnosis of five disea-ses,including atypical endometrial hyperplasia,endometrial polyps,endometrial cancer,endometrial atypical hy-perplasia,and submucous myoma,was slightly higher than that of ResNet18 model,and the AUC of both models was almost above 0.980.②In the binary classification task of accuracy and the evaluation of specificity,the two models were similar,both above 93.00％,and the sensitivity of EfficientNet-B0 model(91.14％)was significantly better than that of ResNet18 model(77.22％).③EfficientNet-B0 model combined with Grad-CAM algorithm could identify the abnormal areas in the image.After biopsy and pathological examination,it was confirmed that about 95％of the marked areas in the model's output heatmap were lesion areas.Conclusions:The hysteroscopy di-agnostic model developed by EfficientNet-B0 model combined with Grad-CAM has high diagnostic accuracy,sen-sitivity,and specificity,and has application value in the diagnosis of endometrial lesions.

Objective:To explore the effectiveness of a hysteroscopic endometrial lesion diagnosis model de-veloped based on deep learning(DL)algorithm combined with gradient-weighted class activation mapping(Grad-CAM)visualization technology.Methods:303 hysteroscopy videos(4781 images)of 291 patients who un-derwent hysteroscopy examination in the Department of Gynecology,Renmin Hospital of Wuhan University from June 1,2021 to December 31,2022 were selected.The dataset was divided into a training set(3703 images)and a test set(1078 images)by weight sampling method.After the training set was used for model learning and train-ing,two model architectures,residual neural network(ResNet18)and efficient neural network(EfficientNet-B0),were selected to verify the model in the test set by five-class and two-class classification tasks,respectively.Tak-ing histopathology as the gold standard,the diagnostic efficacy was evaluated to select the optimal model,and the Grad-CAM layer was embedded in the optimal model to output hysteroscopy images of Grad-CAM.Results:①In the five-class classification tasks,the accuracy of EfficientNet-B0 model(93.23％)was higher than that of Res-Net18 model(84.23％);the area under the curve(AUC)of EfficientNet-B0 model in the diagnosis of five disea-ses,including atypical endometrial hyperplasia,endometrial polyps,endometrial cancer,endometrial atypical hy-perplasia,and submucous myoma,was slightly higher than that of ResNet18 model,and the AUC of both models was almost above 0.980.②In the binary classification task of accuracy and the evaluation of specificity,the two models were similar,both above 93.00％,and the sensitivity of EfficientNet-B0 model(91.14％)was significantly better than that of ResNet18 model(77.22％).③EfficientNet-B0 model combined with Grad-CAM algorithm could identify the abnormal areas in the image.After biopsy and pathological examination,it was confirmed that about 95％of the marked areas in the model's output heatmap were lesion areas.Conclusions:The hysteroscopy di-agnostic model developed by EfficientNet-B0 model combined with Grad-CAM has high diagnostic accuracy,sen-sitivity,and specificity,and has application value in the diagnosis of endometrial lesions.