Maximum Likelihood Probabilistic Model for Pulmonary Embolism Nodule Detection (ML-PPED) using Computer Vision

Computer–aided detection and diagnosis systems have been adopted widely to improve the diagnosis performance by detecting and analyzing the lung diseases. The pulmonary embolism is considered as a fetal condition related to lung where the blood clot cause blockage to the lung arteries and this condition can cause death to the patient. Early detection of blood clot can help to diagnose the pulmonary embolism. In order to detect the PE, lung segmentation and nodule detection is the main task for any CAD system. Several approaches have been introduced to perform the segmentation but the accuracy and false positives of segmentation remains a challenging task in this field. Thus, we focus on the lung segmentation and nodule detection using computer vision approach for PE detection and developed Maximum Likelihood Probabilistic model for Pulmonary Embolism nodule detection (ML-PPED). According to the proposed approach, first of all we extract the lungs regions i.e. left and right lung regions followed by segmentation and finally a maximum likelihood based probabilistic model is developed to detect the lung nodules. The performance of segmentation is measured in terms of dice similarity coefficient and average segmentation error which are computed based on the segmented outcome of the proposed model and ground truth data. The experimental analysis shows that the proposed approach improved the segmentation performance when compared with the existing techniques.