Keywords:
Musculoskeletal system, Radiographers, Ultrasound, Experimental, Computer Applications-Detection, diagnosis, Tissue characterisation
Authors:
P. Conde1, A. Silva1, M. Santos2, A. Silva1; 1Aveiro/PT, 2 Aveiro/PT
DOI:
10.1594/ecr2018/C-2061
Methods and materials
We developed different algorithms based on the Lucas-Kanade optical flow methods [5].
Using a C++ implementation and the OpenCV library,
three different methods were created to quantify the movement of the median nerve in response to wrist flexion and extension based on Optical Flow algorithms: i) ROI tracking,
ii) Point-wise Tracking and iii) Contour Tracking.
Longitudinal and axial acquisitions were taken at approximately the mid-forearm.
ROI Tracking: Consists of a selection of an arbitrary number of points used to calculate the optical flow.
Movement vectors are displayed in real time in each frame of the video.
These vectors represent the movement direction of the best features to track inside the specified region of interest.
Each one of the pre-selected points is assigned a colour depending on its direction,
represented by the compass rose on the left upper corner of Figure 3 (i.e.
if the feature is going on the right direction,
then the red colour will be assigned to it).
This colour assignment is refreshed on every frame.
The total displacement on each direction is then shown on a histogram (Figure 1).
Point-wise Tracking: Consists of a selection of an arbitrary number of points to track.
These points are selected on the first frame and the system will follow their direction.
The direction of each one of these points is also categorized according to the colour scheme of the compass rose on the left upper corner of Figure 2 (i.e.
if the feature is going on the right direction,
then the red colour will be assigned to it).
This colour assignment is refreshed on every frame.
A table with the total displacement and travelled space values are shown on Figure 2 for each point selected.
Contour Tracking: In the Contour Tracking method,
several points are chosen to define a contour which is then tracked.
These points are selected on the first frame and the system will follow their direction.
The direction of each one of these points is also categorized according to the colour scheme of the compass rose on the left upper corner of Figure 3 (i.e.
if the feature is going on the right direction,
then the red colour will be assigned to it).
This colour assignment is refreshed on every frame.
A histogram with the total travelled space values is shown on Figure 3 for each mapped direction.
Total displacement,
total travelled space and initial and final width and height values are shown on Figure 3 and are constantly being calculated.
Displacement and travelled space values are calculated relative to the contour mass centre.
Image Acquisition:
A dynamic ultrasound study of the median nerve in the axial and sagittal plane during flexion and extension of the wrist was performed.
Median nerve was scanned with a high-frequency ultrasound device,
GE Logiq P6 Pro,
with a linear transducer GE 11L at 12 MHz,
with a gain of 66 dB,
at a frame rate of 21 Hz and at 3 cm depth.
The video was exported with a grid of 640x480 pixels per frame.
Longitudinal acquisition started with the patient’s right forearm in supination and hand in neutral,
then the patient performed either maximum wrist flexion or maximum wrist extension (Figure 4).
The images acquired during these movements were then used to test the algorithms.