Machine learning for predicting the risk stratification of 1-5 cm gastric gastrointestinal stromal tumors based on CT.

BACKGROUD: To predict the malignancy of 1-5 cm gastric gastrointestinal stromal tumors (GISTs) by machine learning (ML) on CT images using three models - Logistic Regression (LR), Decision Tree (DT) and Gradient Boosting Decision Tree (GBDT). METHODS: 231 patients from Center 1 were randomly assigned into the training cohort (n = 161) and the internal validation cohort (n = 70) in a 7:3 ratio. The other 78 patients from Center 2 served as the external test cohort. Scikit-learn software was used to build three classifiers. The performance of the three models were evaluated by sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV) and area under the curve (AUC). Diagnostic differences between ML models and radiologists were compared in the external test cohort. Important features of LR and GBDT were analyzed and compared. RESULTS: GBDT outperformed LR and DT with the largest AUC values (0.981 and 0.815) in the training and internal validation cohorts and the greatest accuracy (0.923, 0.833 and 0.844) across all three cohorts. However, LR was found to have the largest AUC value (0.910) in the external test cohort. DT yielded the worst accuracy (0.790 and 0.727) and AUC values (0.803 and 0.700) in both the internal validation cohort and the external test cohort. GBDT and LR performed better than radiologists. Long diameter was demonstrated to be the same and most important CT feature for GBDT and LR. CONCLUSIONS: ML classifiers, especially GBDT and LR with high accuracy and strong robustness, were considered to be promising in risk classification of 1-5 cm gastric GISTs based on CT. Long diameter was found the most important feature for risk stratification.

An Analysis of Computed Tomography Imaging Features and Predictive Factors for Postoperative Recurrence and Metastasis of Abdominal Paragangliomas.

Methods: We studied 51 abdominal PGL patients at the First Affiliated Hospital of Bengbu Medical College, Tongde Hospital, and Sir Run Shaw Hospital, Hangzhou, Zhejiang Province, China, from June 2009 to May 2019. Thereafter, the clinical research data, tumor biomarkers, and CT features were compared between the aggressive PGLs and the nonaggressive PGLs using independent-samples t-tests and chi-square tests. Results: Of the 51 cases, 43 were benign and 8 had malignant tendencies. Postoperative recurrence and metastasis were more likely to occur when the tumor diameter was >8 cm or/and the enhancement degree was not obvious. Clinical symptoms, tumor markers, sex, age, and CT image characteristics including morphology, presence of cystic degeneration, "pointed peach" sign, calcification, hemorrhage, enlarged lymph nodes, and peritumor and intratumor blood vessels were not significantly different between the two groups (p > 0.05). Conclusion: Our findings suggest that CT features, including size >8 cm and enhancement degree, could provide important evidence to assess risk factors for aggressive PGLs.

CT differentiation of gastric ectopic pancreas from gastric stromal tumor.

BACKGROUND: Gastric ectopic pancreas (GEPs) is a rare developmental anomaly which is difficult to differentiate it from submucosal tumor such as gastric stromal tumor (GST) by imaging methods. Since the treatments of the GEPs and GST are totally different, a correct diagnosis is essential. Therefore, we retrospectively investigated the CT features of them to help us deepen the understanding of GEPs and GST. METHODS: This study enrolled 17 GEPs and 119 GST, which were proven pathologically. We assessed clinical and CT features to identify significant differential features of GEPs from GST using univariate and multivariate analyses. RESULTS: In univariate analysis, among all clinicoradiologic features, features of age, symptom, tumor marker, location, contour, peritumoral infiltration or fat-line of peritumor, necrosis, calcification, CT attenuation value of unenhancement phase/arterial phase/portal venous phase (CTu/CTa/CTp), the CT attenuation value of arterial phase/portal venous phase minus that of unenhanced phase (DEAP/DEPP), long diameter (LD), short diameter (SD) were considered statistically significant for the differentiation of them. And the multivariate analysis revealed that location, peritumoral infiltration or fat-line of peritumor, necrosis and DEPP were independent factors affecting the identification of them. In addition, ROC analysis showed that the test efficiency of CTp was perfect (AUC = 0.900). CONCLUSION: Location, the presence of peritumoral infiltration or fat-line of peritumor, necrosis and DEPP are useful CT differentiators of GEPs from GST. In addition, the test efficiency of CTp in differentiating them was perfect (AUC = 0.900).

Short and long-term changes in platelet and inflammatory biomarkers after cryoballoon and radiofrequency ablation.

BACKGROUND: Cryoballoon (CB) versus radiofrequency (RF) ablation response on prothrombotic biomarkers obtained different results at short-term, while long-term is still unknown in atrial fibrillation (AF) treatment. We evaluated short and long-term changes in platelet and inflammatory biomarkers after CB and RF ablation. METHODS: Fifty-eight paroxysmal AF patients were randomized for pulmonary vein (PV) isolation using either CB (n = 29) or RF (n = 29) ablation. Biomarkers of platelet activation [P-selectin (CD62P), CD40 ligand (CD40L), platelet factor-4 (PF-4), mean platelet volume (MPV), platelet-leukocyte ratio (P-LCR), and platelet distribution width (PDW)]; and inflammatory [high sensitivity CRP (hs-CRP) and interleukin-6 (IL-6)] were measured at baseline, 18-24 h and 6-Months postablation. RESULTS: Twenty-four (86.2%) and twenty-six (89.7%) patients remained in sinus rhythm at 6-Months in CB and RF group respectively (p = 0.500). After 18-24 h postablation, CD62P, CD40L, PF-4, hs-CRP, and IL-6 levels were significantly activated in both groups (p < 0.001). However, CD62P was significantly lower in CB than RF (p = 0.017). At 6-Month postablation in CB group, all platelet biomarkers CD62P (p = 0.021), CD40L (p < 0.001), PF-4 (p < 0.001), MPV (p = 0.010), PDW (p = 0.004), and P-LCR (p = 0.033) were significantly decreased compared to baseline levels. However in RF group, CD40L and PF-4 (p < 0.001) significant decreased, CD62P (p = 0.022) increased, and no change in MPV and P-LCR (p > 0.05) compared to baseline levels. hs-CRP and IL-6 levels were comparable between baseline and 6-Months in both groups (p > 0.05). CONCLUSIONS: CB ablation might influence the risk of thromboembolism due to less platelet activation after PV isolation and decreased platelet activation at long-term in maintained sinus rhythm patients compared to RF.

Temozolomide combined with PD-1 Antibody therapy for mouse orthotopic glioma model.

PURPOSE: Temozolomide (TMZ) is the most frequent adjuvant chemotherapy drug in gliomas. PDL1 expresses on various tumors, including gliomas, and anti-PD-1 antibodies have been approved for treating some tumors by FDA. This study was to evaluate the therapeutical potential of combined TMZ with anti-PD-1 antibody therapy for mouse orthotopic glioma model. METHODS: We performed C57BL/6 mouse orthotopic glioma model by stereotactic intracranial implantation of glioma cell line GL261, mice were randomly divided into four groups: (1) control group; (2) TMZ group; (3) anti-PD-1 antibody group; (4) TMZ combined with anti-PD-1 antibody group. Then the volume or size of tumor was assessed by 7.0 T MRI and immunohistochemistry, and the number of CD4 and CD8 infiltrating cells in brain tumor and spleen was evaluated by immunohistochemistry. Western blot was used to evaluate the expression of PDL1. Furthermore, Overall survival of each group mice was also evaluated. RESULTS: Overall survival was significantly improved in combined group compared to other groups (χ2 = 32.043, p < 0.01). The volume or size of tumor was significantly decreased in combined group compared with other groups (F = 42.771, P < 0.01). And the number of CD4 and CD8 infiltrating cells in brain tumor was also obviously increased in combined group (CD4 F = 45.67, P < 0.01; CD8 F = 53.75, P < 0.01). CONCLUSION: Anti-PD1 antibody combined with TMZ therapy for orthotopic mouse glioma model could significantly improve the survival time of tumor-bear mice. Thus, this study provides the effective preclinical evidence for support clinical chemotherapy combined with immunotherapy for glioma patients.