| Abstrakt: |
Separation of diseased brain tissues from normal brain tissues is one of the most important tasks in any system for detecting brain tumors. It is important to note that the field of brain tumor analysis has effectively applied medical image processing techniques, particularly on MR images, to automate crucial procedures including extraction, segmentation, and classification for tumor detection. Computer-aided diagnostic techniques are getting harder to employ and yet have many unresolved issues since tumors can vary in their sizes, locations, and shapes. More research has been centered on the noninvasive imaging capabilities of MRI. As a result, electronic methods are used to diagnose brain malignancies. One of the most popular and commonly used electronic methods for detecting brain cancers is magnetic resonance imaging (MRI). Hence, the MRI brain image based brain tumor model has been designed using the modified Mask R-CNN. In this designed model, Van Cittert algorithm is used for pre-processing the brain tumor MRI images to deblur and improve the quality of the gathered input images. The preprocessed data are segmented using Mask-KRCNN, which divides recognized objects into their pixel-level segments. The KNN classification technique may be used to train the Mask R-CNN algorithm to handle overlapping objects with multiple classes. Accuracy, error, sensitivity, specificity, precision, FPR, FNR, F1-Score, MCC, Kappa, MK, FM, NPV, FOR and FDR are some of the performance metrics for this designed model. The attained performance metrics values for the proposed model are 96, 4, 93, 98, 93, 1, 6, 93, 95, 94, 86, 93, 98, 2, and 2 . Thus, the modified mask R-CNN with KNN algorithm primarily based on segmentation and type for prediction of brain tumor types plays better than the existing version. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink