preview

3D Image Segmentation Paper

Better Essays

At first, the general objective of image segmentation in this project is to aid tumour detection by delineating it from normal brain structures. Segmentation aids to detect, to diagnose, to classify the type and to find the stage; thereby helping the treatment decision. In addition, it also helps to monitor the treatment either chemotherapy or surgery.

Secondly, the human detection and diagnosis is an experience and knowledge drive process. This often leads to variability among experts, especially detecting subtle changes on longitudinal studies. A computational technique, either semi or fully automatic segmentation will aid at various levels of the decision making process. Notably, segmentation faces challenges to match the human perception and …show more content…

At first, the general outline would be to implement a supervised approach. This requires the training dataset that include a cohort of MRI images of normal and Brain tumour. Both will be labelled, trained and tested. We begin with 2D and extended to 3D.

The outline of the stages of the image processing pipeline is; realign (affine), skull scripting, statistical edge detector (multi-scale, multi-modal), extract statistical features, use it to build SVM. The robustness of implementation will be tested with Pixel Correspondence Metric (PCM). The same pipeline will be followed for an unsupervised approach using leave one out (LOOCV). The filter size, parametric and non-parametric features like mean, variance, rank, U Mann Whitney test etc will be evaluated and an optimal performer will be selected.

The training data set will be acquired from Birmingham Children Hospital. In addition, a multi centre collaboration with Children Oncology Group in North America (COG) and Paediatric Brain Tumour Consortium (PBCT) can provide variety, as the paediatric brain tumour has a wide range of histology Louis et

Get Access