학술논문
Target Driven Autoencoder: A Supervised Learning Approach for Tumor Segmentation
Document Type
Conference
Author
Source
2024 International Conference on Emerging Systems and Intelligent Computing (ESIC) Emerging Systems and Intelligent Computing (ESIC), 2024 International Conference on. :273-277 Feb, 2024
Subject
Language
Abstract
A particular kind of neural network called an autoencoder (AE) is trained to discover a reduced and effective representation (encoding) of input data. In most of the applications of AE, it is found that this model has been used for generating the same images in compressed form. AE has also been used for encoding the images and signals. In this work, a supervised autoencoder is designed that is trained with both input images and corresponding targets that are also the images. This allows the model to learn a mapping from the input image to the target image specifically for segmentation. The proposed method is verified and trained with brain MRI images and segmented tumors. The dice score obtained in this method is 98.43% in producing segmented tumors.