IFAS: improved fully automatic segmentation convolutional neural network model along with morphological segmentation for brain tumor detection

• Published in: International journal of information technology (Singapore. Online) 2024-03, Vol.16 (3), p.1517-1525
• Main Authors: Kulshreshtha, Akanksha, Nagpal, Arpita
• Format: Article
• Language: English
• Subjects: Artificial Intelligence; Computer Imaging; Computer Science; Image Processing and Computer Vision; Machine Learning; Original Research; Pattern Recognition and Graphics; Software Engineering; Vision
• ISSN: 2511-2104; 2511-2112
• DOI: 10.1007/s41870-023-01572-5

Image segmentation of brain Magnetic Resonance Imaging (MRI) images is one of the trusted method for tumor detection. The tumors are dynamically grown and can be present in any size and shape. Segmenting the tumor region from tumor affected images with high accuracy is a great challenge. This paper addresses the problem and proposes improved fully-automatic segmentation (IFAS) Convolutional Neural Network model. In IFAS, morphological operation along with a fully- automatic segmentation algorithm on brain MRI images is applied. For the analysis of the fully-automatic segmentation method, the U-net structure is considered for morphological segmentation and the CNN model is used for classification. Three online brain MRI datasets are used for the evaluation and testing of the designed model for brain MRI image classification and segmentation. The proposed model is trained for all three datasets in two stages and testing is conducted for the other two datasets. In the first stage, the model is trained with original MRI images and tested for the other two datasets, with an average dice similarity of 0.707 for dataset 1, 0.7575 for dataset 2, and 0.6063 for dataset 3 being observed respectively. In the second stage, the model is trained on morphologically enhanced images, and then testing is done with the original images of the other two datasets. The average dice similarity observed in the latter case is 0.888 for morphological enhanced dataset 1, 0.791 for morphological enhanced dataset 2, and 0.7197 for morphological enhanced dataset 3. The observed results show that the IFAS model trained on morphologically enhanced images performs better with 99.36% of training accuracy and 98.84% validation accuracy is achieved.