The Value Of Radiomics Model In Predicting Hematoma Expansion In The Early Stage Of Spontaneous Intracerebral Hemorrhage

Objective: To explore the potential role of a radiomics model based on noncontrast computed tomography(NCCT)of the brain in predicting the early expansion of hematoma in patients with spontaneous intracerebral hemorrhage(ICH).Methods: A total of 104 patients diagnosed with spontaneous intracerebral hemorrhage from August 2012 to June 2020 in the First Affiliated Hospital of Wannan Medical College were selected,including 32 males and 72 females,aged(57±11.4)years old,according to the proportion of 7:3,the patients was randomly divided into training set(74 cases)and validation set(30 cases).All patients underwent a head CT scan within 6hours after the onset of the disease,and a re-examination within 24 hours after the baseline CT scan.Hematoma expansion was defined as volume growth exceeding 6 m L or 33% from the baseline volume.,and as the criterion for hematoma expansion,the patients were divided into hematoma enlarged group and non-enlarged group.Export the patient’s baseline NCCT image from the Picture Archiving and Communication Systems(PACS)in DICOM format,use Radi Ant DICM Viewer(64-bit)software to extract the patient’s brain window image from the DICOM image,and then Manually delineate the region of interest(ROI)layer by layer on ITK-SNAP(version 4.6)software,the ROI of each level is merged into a volume of interest(VOI),and the patient VOI information is imported into the AK software developed by GE to extract the feature parameters of radiomics.On the training set,the minimum redundancymaximum relevant(m RMR)is used to remove redundant and irrelevant features,and then the features are selected by LASSO regression,the radiomics model is constructed,and verified on the validation set.The receiver operating curve(ROC)analysis is used to evaluate the differential diagnosis ability of the model,the Hosmer-Lemeshow test is used to evaluate the goodness of fit of the model,and the decision curve is used to analyze the clinical application value and benefit of the evaluation model.Results: A total of 9 radiomic features,including lbp-3D-m1＿firstorder＿Median＿1、lbp-3D-m1＿firstorder＿Mean Absolute Deviation＿1、waveletHHH＿gldm＿Dependence Variance、lbp-3D-m1＿firstorder＿10Percentile＿1、waveletLLH＿firstorder＿10Percentile、wavelet-LLL＿ngtdm＿Contrast、waveletHHH＿glcm＿Cluster Shade、original＿firstorder＿90Percentile、waveletHLL＿glcm＿Contrast,were finally selected to construct a radiomics model;the sensitivity,specificity,and AUC values of the model in the training group are 0.893,0.848,0.910.And the sensitivity,specificity,and AUC values in the validation group are0.909,0.895,0.914.The Hosmer-Lemeshow test shows that the model is well calibrated.Decision curve analysis shows that the model has higher clinical benefits.Conclusion: The radiomics model based on NCCT is more effective in predicting the early expansion of hematoma in spontaneous intracranial hemorrhage,and can provide help for the clinical development of personalized treatment plans.