Monash University
13 files

4D laryngeal CT validation

Download all (9.23 GB)

Data set of 4D laryngeal CT images and inter-arytenoid distance (IAD) measurements. These were collected for the purposes of validating an automated image analysis algorithm. The algorithm consists of a machine learning module which is capable of measuring the 'inter-arytenoid distance' in 4-dimensional dynamic laryngeal CT images. 

The data set includes 8 participants with Parkinson's disease (pwPD) of varying disease durations and severities, and 4 healthy controls. All data was measured by the same investigator to eliminate inter-observer variability. During a continuous CT acquisition scanning period of 5 seconds, participants were instructed to produce five short phonations of /i/ quickly and clearly. Images were re-constructed to produce continuous multiplanar images of the larynx at 100ms per frame. The data set includes the following measurements:

1. mIAD: Manual IAD measurements derived by manually placing a fiducial marker on the most medial aspect of the left and right arytenoid cartilages and calculating the Euclidean distance between them.

2. aIAD: Automated IAD measurements derived by the automated image analysis algorithm. The algorithm would compute the distance between the arytenoid cartilages on every axial slice at every timepoint, with the minimum distance chosen as the IAD.

3. sIAD: 'Same-slice' IAD measurements derived by the automated image analysis algorithm on the same axial slice from which the investigator calculated the mIAD.

The CT images have been stored in .tif format, and are recommended to be opened in an image viewer can view four-dimensional images, such as the open-source ImageJ software. This allows them to be scrolled through in both the z (depth) and t (time) axes. The images have been uploaded at a resolution of 256x256 pixels from the original 512x512 pixels due to file size constraints. Note that the full image set currently stands at a total file size of 9.23gb.