Objective:To develop habitat radiomics models to predict early treatment responses to the hepatic arterial infusion chemotherapy (HAIC) combined with targeted therapy or immunotherapy in advanced hepatocellular carcinoma (HCC) patients, and to guide clinical diagnosis and treatment.Methods:From October 2021 to Decemeber 2023, at Army Characteristic Medical Center of PLA (Chongqing Daping Hospital) and the First Affiliated Hospital of Chongqing Medical University, 94 patients with advanced HCC who received HAIC combined with targeted therapy or immunotherapy were retrospectively enrolled. According to the treatment results, the patients were divided into response group and non-response group. Univariate and multivariate logistic regression were performed to analyze the clinical data of the patients. Based on contrast-enhanced CT images, tumor habitats were delineated and habitat features were extracted with k-means clustering, and the imaging features of arterial and venous phases were also extracted. The least absolute shrinkage and selection operator (LASSO) was used for dimensionality reduction. Feature selection was performed using LASSO to reduce dimensions, and then the selected features were further refined through stepwise logistic regression analysis.Binary logistic regression models were conducted to develop the habitat radiomics model, arterial phase radiomics model (APRM), venous phase radiomics model (VPRM), clinical data model, as well as the combination of radiomics model and clinical data model to predict early treatment (after 2 treatment cycles) response. Receiver operating characteristic curves (ROC) were plotted, and model performance was evaluated by the area under the curve (AUC), calibration curves, and decision curve. The models were validated through Bootstrap methods (1 000 times). DeLong test was used to compare AUC values.Results:The results of cluster analysis identified 3 characteristic habitats in HCC imaging: low-, medium-, and high-enhancement tumor habitats. The proportion of high-enhancement habitats was higher than that in the non-response group. A predictive model was established based on the proportions of these 3 habitats. Based on the proportion of low-, medium-, and high-enhancement habitats within the tumor, a habitat radiomics model was constructed. After LASSO selection and logistic regression analysis, 3 arterial phase and 3 venous phase radiomic features were selected to build the APRM and VPRM, respectively. Logistic regression analysis identified the following factors for the clinical data model: comorbidities ( OR=0.275, P=0.031), maximum tumor diameter ( OR=1.149, P=0.019), red blood cell count ( OR=0.463, P=0.022), alpha fetoprotein >400 μg/L ( OR=3.452, P=0.017), and tyrosine kinase inhibitor therapy ( OR=3.072, P=0.048). Among the single predictive model′s comparison, the AUC of habitat radiomics model was 0.860 (95% confidence interval(95% CI): 0.789 to 0.932), while those of the APRM、VPRM and clinical data model were 0.850 (95% CI: 0.773 to 0.926), 0.855 (95% CI: 0.782 to 0.928), and 0.774 (95% CI: 0.681 to 0.867), respectively, and there were no statistically significant among these models (all P>0.05). Among the combination models, the AUC of the habitat rediomic-clinical data combination model was 0.881 (95% CI: 0.814 to 0.947); the AUC of arterial phase rediomic-clinical data combination model was 0.897 (95% CI: 0.833 to 0.961); and the AUC of venous phase rediomic-clinical data combination model was 0.888 (95% CI: 0.826 to 0.951), but there were no statistically significant among the 3 models (all P>0.05). The calibration curve showed that the habitat rediomic-clinical data combination model had the most accurate predictive probability. Internal validation showed that the AUC of habitat rediomic-clinical data combination model was 0.848 (95% CI: 0.772 to 0.922), and the predictive performance was better than that of the clinical-data model (0.733 (95% CI: 0.670 to 0.863)). Conclusion:The habitat radiomics model based on enhanced CT can effectively predict early treatment responses to the HAIC combined with targeted therapy or immunotherapy in advanced HCC patients, which provides theoretical basis for individualized treatment in advanced HCC.

Objective:To develop habitat radiomics models to predict early treatment responses to the hepatic arterial infusion chemotherapy (HAIC) combined with targeted therapy or immunotherapy in advanced hepatocellular carcinoma (HCC) patients, and to guide clinical diagnosis and treatment.Methods:From October 2021 to Decemeber 2023, at Army Characteristic Medical Center of PLA (Chongqing Daping Hospital) and the First Affiliated Hospital of Chongqing Medical University, 94 patients with advanced HCC who received HAIC combined with targeted therapy or immunotherapy were retrospectively enrolled. According to the treatment results, the patients were divided into response group and non-response group. Univariate and multivariate logistic regression were performed to analyze the clinical data of the patients. Based on contrast-enhanced CT images, tumor habitats were delineated and habitat features were extracted with k-means clustering, and the imaging features of arterial and venous phases were also extracted. The least absolute shrinkage and selection operator (LASSO) was used for dimensionality reduction. Feature selection was performed using LASSO to reduce dimensions, and then the selected features were further refined through stepwise logistic regression analysis.Binary logistic regression models were conducted to develop the habitat radiomics model, arterial phase radiomics model (APRM), venous phase radiomics model (VPRM), clinical data model, as well as the combination of radiomics model and clinical data model to predict early treatment (after 2 treatment cycles) response. Receiver operating characteristic curves (ROC) were plotted, and model performance was evaluated by the area under the curve (AUC), calibration curves, and decision curve. The models were validated through Bootstrap methods (1 000 times). DeLong test was used to compare AUC values.Results:The results of cluster analysis identified 3 characteristic habitats in HCC imaging: low-, medium-, and high-enhancement tumor habitats. The proportion of high-enhancement habitats was higher than that in the non-response group. A predictive model was established based on the proportions of these 3 habitats. Based on the proportion of low-, medium-, and high-enhancement habitats within the tumor, a habitat radiomics model was constructed. After LASSO selection and logistic regression analysis, 3 arterial phase and 3 venous phase radiomic features were selected to build the APRM and VPRM, respectively. Logistic regression analysis identified the following factors for the clinical data model: comorbidities ( OR=0.275, P=0.031), maximum tumor diameter ( OR=1.149, P=0.019), red blood cell count ( OR=0.463, P=0.022), alpha fetoprotein >400 μg/L ( OR=3.452, P=0.017), and tyrosine kinase inhibitor therapy ( OR=3.072, P=0.048). Among the single predictive model′s comparison, the AUC of habitat radiomics model was 0.860 (95% confidence interval(95% CI): 0.789 to 0.932), while those of the APRM、VPRM and clinical data model were 0.850 (95% CI: 0.773 to 0.926), 0.855 (95% CI: 0.782 to 0.928), and 0.774 (95% CI: 0.681 to 0.867), respectively, and there were no statistically significant among these models (all P>0.05). Among the combination models, the AUC of the habitat rediomic-clinical data combination model was 0.881 (95% CI: 0.814 to 0.947); the AUC of arterial phase rediomic-clinical data combination model was 0.897 (95% CI: 0.833 to 0.961); and the AUC of venous phase rediomic-clinical data combination model was 0.888 (95% CI: 0.826 to 0.951), but there were no statistically significant among the 3 models (all P>0.05). The calibration curve showed that the habitat rediomic-clinical data combination model had the most accurate predictive probability. Internal validation showed that the AUC of habitat rediomic-clinical data combination model was 0.848 (95% CI: 0.772 to 0.922), and the predictive performance was better than that of the clinical-data model (0.733 (95% CI: 0.670 to 0.863)). Conclusion:The habitat radiomics model based on enhanced CT can effectively predict early treatment responses to the HAIC combined with targeted therapy or immunotherapy in advanced HCC patients, which provides theoretical basis for individualized treatment in advanced HCC.

Objective To compare the knee function in patients with intraoperative medial collateral ligament(MCL)injury treated with or with?out increased prosthetic constraint. Methods The records of 19 cases who encountered with iatrogenic injury to the MCL during total knee arthro?plasty(TKA)between January 2005 and December 2010 were retrospectively reviewed. Eight patients(LCCK group)were treated with increased prosthetic constraint;the remaining 11 patients(LPS group)received increased prosthetic constraint between January 2005 and December 2010 served as controls. The MCL was repaired after TKA. The complications were observed after operation. Knee society scores(KSS)subjective knee scores were used to assess the knee function. No patient was lost for follow?up. The mean follow?up was 5 years. Results Until last follow?up(60 months),The KSS subjective score was 87.4 for LCCK group compared with 93.3 for the LPS group. No revisions for knee instability were per?formed in the 11 patients treated with non?prosthetic constraint;however,2 patients treated with increased prosthetic constraint were revised due to joint loosening. Conclusion The MCL tear should be repaired during TKA;the type of the prosthesis should not be increased when MCL injury is recognized during TKA.