Insiders Hub

Infant Brain MRI Segmentation

Ashish Sethi

Ashish Sethi

New Delhi, Delhi

1 0
  • 0 Collaborators

Accurate segmentation of infant brain MR images into white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF) in this critical developmental phase are of fundamental importance in studying the normal and abnormal early brain development. ...learn more

Project status: Under Development

Artificial Intelligence

Groups
Student Developers for AI

Intel Technologies
DevCloud, Intel Python, OpenVINO

Code Samples [1]Links [1]

Overview / Usage

​

  • Segmentation of the infant’s brain is difficult as the intensity ranges of voxels in White Matter (WM), Grey Matter (GM), and Cerebrospinal Fluid (CSF) areas are complex and mainly overlapping.
  • Lack of study of an infant’s brain at this stage causes a delay in the detection of the probability of neuropsychiatric disorders in the future.
  • Study at early stages is required for the successful detection of such disorders.
  • We thus aim to make the study of a child’s brain possible by attempting to segment out the complexly overlapping WM, GM, and CSF correctly from their MRI scans using Semantic segmentation algorithms.

Methodology / Approach

  • The dataset is acquired from the iSeg-2019 contest and IBSR
  • T1 weighted images:
    • iSeg-2019: 144 x 192 x 256 (6-month infant dataset)
    • IBSR: 256 x 256 x 128 (Gold Standard adult brain dataset for comparison)
  • iSeg: Total 10 subject images, 7 for training, 2 for validation, 1 for testing
  • IBSR: Total 18 subject images, 10 for training, 4 for validation, 4 for testing
  • Volume input of randomly cropped 80x80x16 is taken from the original depth data.

Labels one-hot encoded into 4 channels representing Background (Back, 1), Cerebrospinal Fluid (CSF, 2), Grey Matter (GM, 3), and White Matter (WM, 4) respectively.

Technologies Used

Pytorch, python, OpenVINO

Repository

https://github.com/geekysethi/mri_segmentation

Comments (0)