Research On Computer-Aided Automatic Evaluation Algorithm Of Collateral Circulation In Cranial CTA

Cerebrovascular disease is the leading cause of death among Chinese residents,and the lifetime risk of stroke in the Chinese population is the highest in the world.Stroke is one of the main causes of disability and death,and it can be divided into hemorrhagic stroke and ischemic stroke.Ischemic stroke is more common,and about 8 out of 10 stroke patients suffer from ischemic stroke.In clinical practice,doctors diagnose stroke by using computed tomography angiography(CTA)image to accurately evaluate the collateral circulation in stroke patients.This imaging information is of great significance in assisting clinical doctors to determine the patient’s treatment plan and prognosis.Currently,great progress has been made in the field of computer-aided diagnosis technology in medicine by using artificial intelligence,especially in the evaluation of collateral circulation based on machine learning for cranial CTA.However,research in this field has not yet explored the combination of clinical features and radiomics features to evaluate collateral circulation.Meanwhile,in related research based on deep learning algorithms,researchers usually only use single-phase data for training,lacking the temporal dimension information of multi-phase image data.This makes it difficult for the model to learn more comprehensive and effective collateral circulation feature representation,thereby limiting its performance.Therefore,combining data for training is expected to improve the accuracy and reliability of collateral circulation evaluation.In this study,artificial intelligence technology was mainly used in combination with CTA to evaluate collateral circulation,and the following research work was carried out from this perspective.1.Generating cranial CTA image dataset.In response to the current lack of public datasets,two datasets were generated to explore the image features of collateral circulation in patients.Multiple-phase CTA sequence images were selected,preprocessed,and augmented to alleviate the problem of insufficient training samples.Firstly,a multi-phase cranial CTA image dataset was generated,followed by the selection of eligible ischemic stroke patients and the labeling of their regions of interest to extract radiomics features.The second dataset,Radiomics Clininc CTA,was constructed by combining clinical data with radiomics features.Research will be conducted on both datasets.2.Research on collateral circulation evaluation algorithm of fusion data based on radiomics.In order to address the issue of precise evaluation of collateral circulation in ischemic stroke patients,two collateral circulation evaluation models were designed from the perspective of data-level fusion and feature-level fusion.To remove redundant features in the dataset,Levene and T-tests were used for feature pre-selection.Then,feature dimensionality reduction was performed using LASSO and random forest algorithms.Multiple machine learning algorithms were used for classification model training on the data-level fusion data after feature engineering processing.The optimized data-level fusion model was obtained through experiments on the Radiomics Clininc CTA dataset,with an accuracy and AUC value both exceeding 86%.The feature-level fusion classification model was then trained and tested,and it was found that its performance was slightly better than the optimized model of data-level fusion classification.Moreover,experimental comparisons showed that fusion datasets can better distinguish the features of good and poor collateral circulation compared to radiomics datasets.3.Research on Attention-Based Collateral Circulation Evaluation Algorithm.We propose an effective hybrid mechanism to assist the feature encoding network in evaluating the degree of collateral circulation in the brain.By using a hybrid attention mechanism,additional guidance and regularization are provided to enhance the collateral circulation feature representation across multiple stages.Time dimension information is added to the input,and multiple feature-level fusion modules are designed in the multi-branch network.The first fusion module in the single-stage feature extraction network completes the fusion of deep and shallow vessel features in the single-branch network,followed by the multi-stage network feature fusion module,which achieves feature fusion for four stages.Tested on a dataset of multi-stage cranial CTA images,the accuracy and AUC value both exceeding 87%.The experimental results demonstrate that the addition of these modules can fully explore collateral vessel features,improve feature expression capabilities,and optimize the performance of deep learning network model.