Enhanced Pulmonary Embolism Detection in CT Angiography Using Spectral ResNet Hyper Convolutional Neural Network

Blood clot in a lung blood artery causes pulmonary embolism (PE). This is diagnosed via Computed Tomography and Angiography (CTPA) scans.CT pulmonary angiography-based pulmonary embolism (PE) detection is one of the most crucial imaging modalities for risk assessment and disease severity evaluation in these patients.The main objective of this study is to detect pulmonary embolism using deep learning methods. Deep learning-based Computed Tomography Angiography (CTA) can detect acute Pulmonary Embolism (PE) because rapid recognition and prompt treatment can significantly reduce the risk of death. DL methods can effectively detect blood clots, but the previous methods make it difficult to detect PE with a single scan manually cannot classify the complex Features, and reduces accuracy, precision, recall rate, and detection of the images. Initially, the Bio-inspired Adaptive Gaussian filter technique was to remove additional noise and potential holes within the blood clot domain during pre-processing. The Semi-Autonomous Cascading Segmentation (SACS) algorithm is used to identify the clot region for measuring changes in ventricular and clot volumes. Additionally, the Support Scalar Vector Feature (SSVF) technique can identify crucial features in the blood clot data and perform vector margin analysis. A Spectral ResNet Hyper Convolutional Neural Network (SRHCNN) method is designed for classification to identify the Blood clot region to support hemodialysis.The accuracy and effectiveness of pulmonary embolism identification are increased by using Resnet-Net-CNN-based pulmonary embolism detection. The proposed model shows promising results in accurately detecting PE cases from CT images, with high accuracy in sensitivity, and specificity, and improved detection efficiency.