ABSTRACT
Tumor is a dangerous disease that kills thousands of people every day in India. The key cause is abrupt organ mass development. Tumors are divided into two types: benign and malignant. The main issue is that everyone’s size and shape are different. In order to discover this at an early stage, perfect segmentation is critical. Since manual segmentation varies between observers, it is sometimes inconclusive. This work introduces a novel deep neural network-based methodology for segmenting brain tumors and their sub-regions from multimodal MRI images, as well as survival prediction using characteristics generated from segmented tumor sub-regions and clinical characteristics. This study defines a computer-aided diagnostic approach. This section defines a computer-aided diagnostic approach that uses a deep learning-based technique called TransResV-Net, which is a V-net extension, to segment tumor cells. The revised V-Net outperformed the current finest practice strategies. V-net is more successful than the traditional network for pixel-based segmentation, and it is one of the 90simplest methods for segmenting tiny data sets. This methodology is used to partition usable data collecting. This model is applied to the usable BRATS 2015, Kaggle data set comprising 120 patients’ MRI images to apply the segmentation. Apart from this, various databases are available. But this BRATS 2015 is chosen for this experiment. This experiment is being conducted for 20 epochs, and the result is being compared with the supplied ground truth, and the segmented outcome is assessed using metrics such as intersection-over-union; the dice coefficient was reported as 0.72 and 0.87, respectively. As compared to well-known and benchmark approaches, the result demonstrated a significant difference. Researchers in the fields of bioinformatics and medicine will benefit greatly from the high-performing TransResV-Net.
