Locating Vital Thoraco-Abdominal Organs for Body Armour Design: Converting MRI to CT using Convolutional Neural Networks


The input (left) and output (middle) from the second self-trained model, compared to the output (right) from the first self-trained model.


This project is an example of applying an existing technology (deep learning) into a new area (abdominal medical imaging), which can serve a unique purpose which is soldier body armour classification.

Currently, body armour does not account for different body shapes, and so cannot always protect the torso organs. New body armour designs require the correct position of the torso organs in the standing position.

To correctly identify and position these organs (segmentation) is the overarching desire of this research, which requires the use of an algorithm called Orthovis (organ segmentation) and Computed Tomography (CT) data.

Previous validation of the method used the CT of a phantom model of the human torso, and organ segmentations using Orthovis. The next stage was to validate using human data.

The in vivo human data provided to this research is Magnetic Resonance Imaging (MRI); thus, converting the data to CT became necessary. The first sep of MRI to CT conversion using convolutional neural networks has been developed, however more refinement of the MR to CT conversion is required. The CT image will then be used for identifying organ location (segmentation).




Body armour

organ segmentation

convolutional neural networks

image processing

medical imaging

Updated:  1 June 2019/Responsible Officer:  Dean, CECS/Page Contact:  CECS Marketing