MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

A Fuzzy C-Means Multi-Parametric Segmentation Method for Temporal Stability Area Definition On Shear Wave Elastography Image Sequences

I Gatos1 , S Tsantis2 , P Zoumpoulis3 , I Theotokas4 , D Mihailidis5 , J Hazle6 G Kagadis7*, (1) University of Patras, Rion, Ahaia, Greece (2) University of Patras, Rion, Ahaia, Greece (3) Diagnostic Echotomography S.A., Kifissia, Attiki, Greece (4) Diagnostic Echotomography S.A., Kifissia, Attiki, Greece (5) University of Pennsylvania, Philadelphia, PA, (6) UT MD Anderson Cancer Center, Houston, TX, (7) university Patras, Rion - Patras, Greece

Presentations

(Sunday, 7/29/2018) 4:00 PM - 5:00 PM

Room: Exhibit Hall | Forum 1

Purpose: To define a reliable, temporally stable area in Shear Wave Elastography (SWE) image sequences using texture and wavelet coefficient information fed to a Fuzzy C-Means clustering algorithm.

Methods: The clinical dataset includes 200 subjects (88 Healthy and 112 with Liver Biopsy validated Chronic Liver Disease). Each subject had an Ultrasound (US) SWE Abdomen examination performed on Aixplorer (Supersonic Imagine) US device. From each examination, 4 SWE images of the same liver area, with 2 seconds time distance between each other, were extracted towards temporal stability estimation. For each set of the 4 images, an automatic SWE color-box detection algorithm was employed to extract the elastographic information and reduce processing time. For each extracted color-box an RGB to Stiffness process was implemented converting RGB values to the corresponding Stiffness values according to Aixplorer's provided color-bar. For each stiffness-box, the Dyadic Wavelet Transform (DWT) "Atrous" algorithm was calculated. Subsequently, the first stiffness-box and its corresponding DWT box are subtracted from the other three stiffness and DWT boxes respectively, to acquire 6 image-boxes that are related to temporal stability. These 6 image-boxes are fed to a Fuzzy C-Means clustering algorithm to provide 2 clusters of high and low temporal stability. Mean Stiffness Temporal Standard Deviation (mTSD) was calculated for evaluation of Temporal Stability.

Results: Segmentation results showed that stiffness temporal standard deviation on the masked was significantly lower than the unmasked (whole) stiffness-box (mTSDmask = 3.6±1.2, mTSDwhole = 6.8±3.7).

Conclusion: The proposed algorithm can define areas of SWE images with temporal stability and could be used by clinicians for quick reliable stiffness measurement estimation. It could also be used by Computer Aided Diagnosis (CAD) systems for reliable feature extraction and selection for a more accurate CLD diagnosis.

Keywords

Segmentation, CAD, Classifier Design

Taxonomy

IM- Ultrasound : Shear wave

Contact Email