Efficient diagnosis of adenomyosis medical condition using automated diagnosis model over uterine MR images

Abstract

Adenomyosis is a medical condition where the uterus inner lining i.e. endometrium breaks through the myometrium or uterus muscle wall. The task of classification is considered as a major challenge in the field of medical imaging for the diagnosis of adenomyosis medical condition. So far, there are no methods adopted for classifying the adenomyosis condition of women. Since, the adenomyosis is a non-cancerous type of lesion, detection of this condition in MR images requires precise measurements. This thesis focus on two different methods: GSO-FCM-GLCM and BGSO-FCM-GLCM-Entroxon to improve the task of classification of malignant and non-malignant regions of women uterus for the detection of adenomyosis. Prior to classification, it involves various pre-processing steps to improve the task of accurate classification of women uterus to detect the adenomyosis condition. Two different approaches have proposed in this study to present the MR images initially with preprocessing operations for the removal of impulse noise. After the removal of impulse noise, the less contrast nature of MR image is improved using Enhanced Progressive based Equalization. The images are then segmented into various regions using FCM algorithm, GSO based FCM algorithm and BGSO based FCM algorithm. The GLCM and GLCM-Entroxon algorithm is used to extract the features from the segmented image. An IWD algorithm selects the optimal features, where FNN and BPNN is used for classifying the adenomyosis medical condition. The performance of proposed methods and existing method against different performance metrics are tested using an experimental study. newline