,
Ph.D. Research Scholar, Erode Arts & Science College , Erode, Tamil Nadu , India
,
Associate Professor & Head of Department, Erode Arts & Science College , Erode, Tamil Nadu , India
Assistant Professor, PSG College of Arts and Science , Coimbatore, Tamil Nadu , India
Proper brain tumor segmentation is of great importance in the diagnosis and treatment planning. In this paper, the author presents an EfficientNet-modified U-Net-based system to segment the glioma in the pre-operative MRI scans using the BraTS 2018 dataset. This data consists of four types of MRI (T1, T2, T1Gd, and FLAIR). The model uses the EfficientNet-B6 as a tool to enhance the accuracy of feature extraction and segmentation by striking a balance between depth, width, and resolution in the network. The preprocessing of data is done by performing image transformation, subset division, and feature scaling, followed by feeding the data into the U-Net structure. The performance attributes of the model were assessed based on common measures such as Intersection over Union (IoU), precision, and recall. The presented model yielded an IoU of 92.7%, which is higher than other approaches, which got an IoU range of 81-91% on the BraTS dataset. Moreover, the accuracy and the recall of the model were 91.5% and 92.7%, respectively, which is by far better than other models like CAE (83.6 %, 83.2 %) and FCA (85.5%, 84.7 %). These findings indicate that the EfficientNet-enhanced U-Net architecture is effective in the accurate segmentation of brain tumors, and the diagnostic accuracy is improved. The model not only saves the cost of computations by utilizing the feature extracting capabilities of EfficientNet, which is efficient, but also offers automated support to clinicians, reducing human error. The solution is a powerful, scalable method to brain tumor segmentation in clinical practice, which helps in making a correct diagnosis and treatment decision-making.
This is an open access article distributed under the Creative Commons Attribution Non-Commercial License (CC BY-NC) License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
0
The statements, opinions and data contained in the journal are solely those of the individual authors and contributors and not of the publisher and the editor(s). We stay neutral with regard to jurisdictional claims in published maps and institutional affiliations.