Pseudo Computed Tomography Image Generation from Brain Magnetic Resonance Image for Radiation Therapy Treatment Planning Using DCNN-UNET

Abstract

Relative to Computed Tomography (CT), the increased soft tissue contrasts of magnetic resonance imaging (MRI) makes it a suitable imaging method to decide radiation therapy (RT). When MRI scans are used for therapy planning, a CT scan is still required for dosage calculation and x-ray-based patient placement. This raises workload, leads to incertitude owing to the requisite of image registration inter-modality and requires needless irradiation. Even though it would be advantageous to only use MR images, a way of estimating a pseudo-CT (pCT) must be used to generate electron density mapping and patient reference imagery. So, this paper brings an effective deep learning model to generate synthesized CT from MRI images using the following steps; a) data acquisition where CT and MRI scan images are collected, b) preprocessing of images to avoid the anomalies and noises using techniques like outlier elimination, data smoothening and data normalizing, c) feature extraction and selection using Principle Component Analysis (PCA) & regression method, d) generating pCT from MRI using Deep Convolutional Neural Network and UNET (DCNN-UNET). Further, we assessed metrics such as DC, SSIM, MAE and MSE for this model. However, our suggested model outperforms with an accuracy of 95%.