Anita Desiani, Deshinta Arrova Dewi, Dite Geovani, Muhammad Arhami, Emy Setyaningsih, Lucky Indra Kesuma
Brain tumors are among the most dangerous cancers and have a high death rate. One way to deal with the high rate of brain tumor cases is through early detection of brain tumors on brain MRI images. Early detection of tumors on brain MRI images is carried out by detecting abnormalities in nonenhancing tumor cells, peritumoral edema, and enhancing tumor cells on brain MRI images. This can be achieved through segmentation to obtain the detailed features of these cells on brain MRI images. The CNN architecture that is widely used in three-dimensional image segmentation is the volumetric UNet (V-UNet) architecture. This study proposes the VB Light-UNet architecture to segment nonenhancing tumor, peritumoral edema, and enhancing tumor cell features on brain MRI images. The VB Light-UNet architecture is a V-UNet architecture that removes the bridge part, adds batch normalization at each convolution layer, and uses pooling indices in the max pooling operation. This study still maintains the three-dimensional MRI image without converting it to a two-dimensional image first, to utilize all spatial information in the three-dimensional image. Evaluation of the performance of the VB Light-UNet architecture in brain tumor segmentation on brain MRI images showed accuracy, sensitivity, MCC, and G-mean above 0.92, specificity and Dice score above 0.83, IoU of 0.74, and Cohen’s kappa of 0.52. The segmentation results obtained are very similar to the ground truth. These results show that the VB Light-UNet architecture is robust and capable of precisely and accurately segmenting brain tumors on brain MRI images. Copyright © 2026 Anita Desiani et al. Applied Computational Intelligence and Soft Computing published by John Wiley & Sons Ltd.
Department of Mathematics, Universitas Sriwijaya, South Sumatera, Indralaya, Indonesia; Department of Computer Science and Technology, INTI International University, Negeri Sembilan, Putra Nilai, Malaysia; Department of Engineering Science, Universitas Sriwijaya, South Sumatera, Indralaya, Indonesia; Department of Informatics Engineering, Politeknik Negeri Lhokseumawe, Aceh, Lhokseumawe, Indonesia; Department of Computer Systems Engineering, Akprind University, Special Region of Yogyakarta, Yogyakarta, Indonesia; Department of Electrical Engineering, Universitas Sriwijaya, South Sumatera, Indralaya, Indonesia
Research at a Glance
Register to unlock
Topics & SDG Alignment
Register to unlock
Collaboration
Register to unlock
Author Profile (Selected)
Register to unlock
References Overview
Register to unlock
Journal & Source
Register to unlock
Metadata & Integrity
Register to unlock