Brain Tumor Segmentation in MRI Imaging Using an Efficient UNet LITE Model

Abstract

Brain tumor detection is a critical task in medical image analysis, as early and
accurate diagnosis plays an important role in effective treatment planning and improving
patient survival rates. Magnetic Resonance Imaging (MRI) is widely used for identifying
brain abnormalities, but manual analysis by radiologists can be time-consuming and prone to
human error. In existing systems, Convolutional Neural Networks (CNN) have been used for
tumor detection; however, they often face limitations in precisely identifying tumor
boundaries. To address this issue, this study proposes a brain tumor segmentation approach
using a lightweight U-Net architecture (UNet-Lite). The proposed model is designed to
efficiently segment tumor regions from MRI images by utilizing an encoder–decoder
structure with skip connections that preserve spatial information and improve segmentation
accuracy. The system is trained and evaluated on a dataset of brain MRI images to
automatically detect and highlight tumor regions. Experimental results demonstrate that the
proposed UNet-Lite model provides improved accuracy and better segmentation performance
compared to traditional CNN-based approaches. The automated system can assist radiologists
by providing faster and more reliable tumor detection, reducing diagnostic errors and
enhancing clinical decision-making. This work highlights the effectiveness of deep learningbased
segmentation techniques in medical imaging and their potential to support early
diagnosis and treatment of brain tumors