Purpose To explore the value of deep learning feature prediction based on the ResNet50 deep residual network model for predicting early hematoma expansion in spontaneous intracerebral hemorrhage without traditional imaging manifestations. Materials and Methods A retrospective study was performed on 235 patients with spontaneous intracerebral hemorrhage in Jiangdu People's Hospital Affiliated to Yangzhou University from January 2019 and December 2022. These patients had undergone their initial plain cranial CT scan within 6 hours of symptom onset and a subsequent follow-up scan within 24 hours of admission. They were randomly assigned to a training set consisting of 188 cases and a test set of 47 cases,using an 8︰2 ratio. The region of interest (ROI) of hematoma was traced layer by layer on the first plain head CT,and image genomics features were extracted. The maximum two-dimensional cross-sectional ROI of the hematoma 3D-ROI,as well as ROI images at 1 mm and 2 mm above and below the maximum two-dimensional cross-sectional ROI,were then cut and input into the pre-trained ResNet50 model for feature extraction. The image genomics features were then fused with the extracted deep learning features using a least absolute shrinkage and selection operator regression model. A support vector machine classifier was used to construct a prediction model,which was evaluated using receiver operating characteristic curves and decision curve analysis. Results In the training set,the area under curve (AUC) of the deep learning feature model was 0.972,which was higher than that of the image genomics feature model (0.951) and the fused feature model (0.968),but this difference was not statistically significant (P>0.05). In the testing set,the AUCs of the deep learning feature model and the fused feature model were 0.867 and 0.895,respectively,which were significantly higher than that of the image genomics feature model (0.833),with statistically significant differences (Z=-1.794,-2.191,both P<0.05). The AUC of the fused feature model showed an improvement compared to the deep learning feature model,but the difference was not statistically significant (P>0.05). In the test set,decision curve analysis revealed that the fused feature model yielded greater benefits compared to both the deep learning feature model and the radiomic feature model. Conclusion The deep learning feature model based on ResNet50 deep residual network shows better performance in predicting early hematoma expansion than the image genomics feature model,and the fused feature model has a beneficial effect on predicting hematoma expansion. This deep learning approach provides a prediction tool with supervisory capability for clinical decision-making.

Purpose To explore the value of deep learning feature prediction based on the ResNet50 deep residual network model for predicting early hematoma expansion in spontaneous intracerebral hemorrhage without traditional imaging manifestations. Materials and Methods A retrospective study was performed on 235 patients with spontaneous intracerebral hemorrhage in Jiangdu People's Hospital Affiliated to Yangzhou University from January 2019 and December 2022. These patients had undergone their initial plain cranial CT scan within 6 hours of symptom onset and a subsequent follow-up scan within 24 hours of admission. They were randomly assigned to a training set consisting of 188 cases and a test set of 47 cases,using an 8︰2 ratio. The region of interest (ROI) of hematoma was traced layer by layer on the first plain head CT,and image genomics features were extracted. The maximum two-dimensional cross-sectional ROI of the hematoma 3D-ROI,as well as ROI images at 1 mm and 2 mm above and below the maximum two-dimensional cross-sectional ROI,were then cut and input into the pre-trained ResNet50 model for feature extraction. The image genomics features were then fused with the extracted deep learning features using a least absolute shrinkage and selection operator regression model. A support vector machine classifier was used to construct a prediction model,which was evaluated using receiver operating characteristic curves and decision curve analysis. Results In the training set,the area under curve (AUC) of the deep learning feature model was 0.972,which was higher than that of the image genomics feature model (0.951) and the fused feature model (0.968),but this difference was not statistically significant (P>0.05). In the testing set,the AUCs of the deep learning feature model and the fused feature model were 0.867 and 0.895,respectively,which were significantly higher than that of the image genomics feature model (0.833),with statistically significant differences (Z=-1.794,-2.191,both P<0.05). The AUC of the fused feature model showed an improvement compared to the deep learning feature model,but the difference was not statistically significant (P>0.05). In the test set,decision curve analysis revealed that the fused feature model yielded greater benefits compared to both the deep learning feature model and the radiomic feature model. Conclusion The deep learning feature model based on ResNet50 deep residual network shows better performance in predicting early hematoma expansion than the image genomics feature model,and the fused feature model has a beneficial effect on predicting hematoma expansion. This deep learning approach provides a prediction tool with supervisory capability for clinical decision-making.

Objective To observe the value of subregional non-contrast CT(NCCT)radiomics features based on habitat imaging technology for predicting hematoma expansion(HE)in patients with spontaneous intracranial hemorrhage(sICH).Methods Data of 228 sICH patients with negative conventional imaging signs were retrospectively analyzed and divided into HE group(n=99)or non HE(NHE)group(n=129)based on the occurrence of HE nor not.also divided into training set(n=182)or test set(n=46)at a ratio of 8:2.Clinical data,NCCT data and laboratory examination results were compared between groups.Logistic regressive analysis was performed to screen the impact factors of HE.ROI of whole hematoma(ROIwhole)was sketched and clustered into 3 sub-regions(ROIsub1,ROIsub2 and ROIsub3,the latter located in the critical area between hematoma and brain tissue)with habitat imaging technology,and radiomics features of ROI were extracted and screened.Then 4 prediction models were constructed based on the above 4 ROI,and the efficacy of each model for predicting HE was analyzed.Results The fasting blood glucose in HE group was higher than that in NHE group(t=2.047,P=0.041),which was not independent impact factor for predicting HE in sICH patients(P=0.070)according to logistic regression analysis.The area under the curve of ROIsub3 radiomics model for predicting sICH HE in training and test set was 0.945 and 0.863,respectively,not significantly different with that of ROIwhole(0.921,0.813),ROIsub1(0.925,0.807)nor ROIsub2(0.909,0.720)(all P＞0.05).Decision curve analysis showed that ROIsub3 radiomics model could bring greater benefits than the other 3 models.Conclusion NCCT radiomics features of the critical area between hematoma and brain tissue based on habitat imaging technology had high value for predicting HE in sICH patients.

Objective:To investigate the clinical significance of fluid-attenuated inversion recovery (FLAIR) vascular hyperintensities (FVHs) in patients with chronic atherosclerotic middle cerebral artery occlusion.Methods:From July 2016 to November 2020, patients admitted to the Department of Neurology, Jiangdu People's Hospital of Yangzhou and first found chronic atherosclerotic middle cerebral artery occlusion were enrolled. The demographic, clinical and MRI imaging data were collected. According to the presence or absence of acute cerebral infarction, they were divided into the non-acute cerebral infarction group and the acute cerebral infarction group. According to the modified Rankin Scale score at 3 months after the onset, patients with acute cerebral infarction were further divided into the good outcome group (≤2) and the poor outcome group (>2). A multivariate logistic regression model was used to analyze the independent correlation between FVHs and the risk of cerebral infarction in patients with chronic atherosclerotic middle cerebral artery occlusion and the outcome in patients with cerebral infarction. Results:A total of 94 patients with chronic atherosclerotic middle cerebral artery occlusion were enrolled, including 38 with non-acute cerebral infarction (40.4%) and 56 with acute cerebral infarction (59.6%). In patients with acute cerebral infarction, 13 (23.2%) had a poor outcome, and 43 (76.8%) had a good outcome. The fibrinogen level, the proportion of patients with FVHs and the FVHs score in the cerebral infarction group were significantly higher than those in the non-cerebral infarction group (all P<0.05). Multivariate logistic regression analysis showed that the FVHs score was significantly, independently, and positively correlated with the risk of acute cerebral infarction (odds ratio 2.524, 95% confidence interval 1.400-4.552; P=0.002). The National Institutes of Health Stroke Scale score at admission, the modified Rankin Scale score at admission, and FVHs score in the poor outcome group were significantly higher than those in the good outcome group (all P<0.05). Multivariate logistic regression analysis showed that there was a significant independent negative correlation between the FVHs score and the outcome of patients with cerebral infarction (odds ratio 0.144, 95% confidence interval 0.045-0.459; P=0.001). Conclusions:FVHs suggest that the blood supply is in a state of decompensation. When FVHs are present, the risk of cerebral infarction in patients with chronic middle cerebral artery occlusion is significantly increased; the wider the range of FVHs, the higher the risk of poor outcome after cerebral infarction.