Target Outcome

The goal of this data analysis specification is to provide tissue thicknesses of skin, fat and muscle for both in vivo and in vitro experimentation. The output data can then be used for modeling purposes, statistical data analysis, and comparison to MRI and CT thicknesses.

Prerequisites

Infrastructure

Infrastructure/InstrumentedUltrasound

Previous Protocols

Requirements

Python [Version 2.7]

Procedures

Ultrasound Thickness Analysis

Thickness Analysis Software

Version 1.0 places red circles at the coordinates of the previous frame that was analyzed, starting with the second frame. Version 1.1 moves red circles to the location with the closest matched distribution to the coordinates in the previous frame, starting with the second frame. Version 2.0 does not require the user to hit 'enter' to save the thicknesses and allows the user to cycle through all of the locations for one subject instead of restarting the program after each location.

Setting up the Directory
1. Download and unpack the desired Subject ID folder, make sure you use the 'associated' data if you do not want to associate the files yourself.
  Note: If you have just completed the file association, confirm that the dicom and tdms files within the Ultrasound and Data folders have been matched (should begin with a three digit number) and that the corresponding deltaT values (as xml files) are located in the TimeSynchronization folder.
The folder/file configuration should be the following
- MULTIS###-# (Subject folder containing all data - the required data for the thickness measurements is shown below)
  - Configuration - Contains subject xml file for accepted trials
  - Data - Contains tdms files
  - Ultrasound - Contains dicom files
  - TimeSynchronization - Contains xml files with deltaT values for matched trials
Required Python scripts
- TA_sfm_v#.#.py
- tdsmParserMultis.py
- XMLparser.py
- Plot_thicknessForce.py
Running the manual thickness measurement script
1. Install libraries as needed
2. Change the directory to match where the Python scripts are located
3. Run script
  - The first dialog window will allow you to select the Subject folder (you may also choose to hard code the path to the Subject folder, found within the getFiles() function)
    - Browse through to find the location of this folder and click 'ok'
  - A list will appear in another window which will show all the trials for that subject (you may have to scroll down to see the entire list)
  - Double-click on the location you would like to analyze
  - After a couple of seconds, the first image will appear
    - The program behaves differently depending on if the trial is anatomical or indentation
      - Anatomical - The program begins with the frame corresponding to the lowest resultant force
      - Indentation - The program begins with the frame that corresponds to the beginning of indentation
  - Drag the red dots so that the center of the dot is over each boundary (use the built in matplotlib toolbar to zoom in if needed)
    - Transducer/Skin boundary
    - Skin/Fat boundary
    - Fat/Muscle boundary
    - Muscle/Bone boundary
  - When you are happy with your selection, press 'next image' and the results will be calculated and saved to an xml file.
  - The next frame for analysis will automatically appear
    - Anatomical trials will give you the image corresponding to the next lowest force, until all frames within the first pulse and last pulse range have been analyzed
    - Indentation trials will give you the next image in the sequence until reaching the maximum force
  - After analyzing the frames for that location, hit the 'Done' button and the list of images for that subject will re-appear and you can analyze a new location.
  - The xml file that was saved will be re-formatted using pretty print (only if you hit the 'Done' button)
  - A PNG folder within the Analysis folder will be created that has saved images of the first frame that was analyzed as well as a graph showing the force/thickness relationships
  - You may end the program at any time by clicking the 'X' button in the upper right hand corner.
    - All windows will be closed and the script will stop running
Data storage
- The xml file will be named using the tdms filename with '_manThick'+ time stamp appended
  - Version of software used for analysis (TA_sfm.py)
  - File Info includes the time stamp at the beginning of analysis as well as the user name at the time of program execution
  - Subject shows the MULTIS ID name
  - Source shows the Filename of the data (tdms) file
- A new 'Frame' child is appended for each frame that is analyzed
  - The Frame child includes the following information:
    - Value - Frame number
    - Time - Reference time from the ultrasound time vector data (in ms)
    - Forces - Forces in the x, y, and z directions (in N)
    - Moments - Moments in the x, y, and z directions (in Nm)
    - Thicknesses - Skin, Fat, and Muscle layers (in mm)

Thickness Inclusion XML

A single XML document was also developed to streamline the process of searching for the appropriate thickness XMLs. This file is used for data analysis scripts as well as modeling scripts.

Repeatability & Reproducibility

Bland & Altman plots (example below) showing the thicknesses of the first frame of each location that was analyzed (Plot_thicknessForce_reliability_ALL.py)
- Blue line shows the mean difference
- Red lines show 2 standard deviations above and below the mean difference
- Script outputs the trial number of the differences that are further than 2 standard deviations from the mean difference.
Results for reproducibility of MULTIS004 (Rici and Ahmet both analyzed all anatomical trials). Bland Altman Plot

Bland Altman Plot Example

Automated Thickness Measurement Development

Note: The automated development has been suspended until data is collected on additional subjects.
Texture analysis of boundaries (transducer/skin, skin/fat, fat/muscle, and muscle/bone)
- Extract rectangles (3mm x 6mm) with boundary located in center
- Analyze by regions
  - Fit to a Gaussian curve (average horizontally and then normalize by peak intensity)
  - Examine moments of the image using skimage.measure module (Hu's moments, first invariant is analogous to moment of inertia)

Initial results report (using 12 locations from MULTIS-001-1): Specifications/DataAnalysis/report1.pdf

Preliminary Results/Questions

In Vivo Thickness Analysis

Report for radiologist examination. Thickness Analysis Questions for Radiologist
- Meeting notes with Dr. Delzell (2016-11-21)
  - Images with unsure bone boundaries were difficult to determine if it was bone or fascia, Dr. Delzell suggests using the transverse imaging mode to find the bone first and then switching to the longitudinal mode to capture our images.
  - When looking for the boundary of fat to muscle, the fascia of the muscle should be clearly defined by a bright signal (if in the correct plane)
  - Many of the distal locations have little to no fat
  - See Thickness Analysis Comments from Dr. Delzell for more comments relating to specific images.

In Vitro Ultrasound/MRI Thickness Comparison

Trial 1

MRI: Trial 4, Ultrasound: MULTIS_C_008-1
Used ITK snap to estimate MRI thicknesses
There are quite a few discrepancies between the two imaging modalities
- Initial comparison of US and MRI
- Initial thoughts:
  - Consistency in position of leg between Ultrasound and MRI - Potentially a stand for the head of the femur
  - Positions (proximal, central, distal, medial, lateral, anterior, posterior) from the MRI may not match the positions of the Ultrasound - this will be improve once Optotrak is implemented
  - Concern with consistency between in vivo and in vitro experimentation - Tyler will work with Tammy to learn the in vivo protocol and apply it to in vitro
  - Improvements needed:
    1. Consistency between ultrasound of in vivo and in vitro
    2. Consistency between ultrasound and MRI of the in vivo extremities

Trial 2

CMULTIS002-2 (Leg)
- Comparison of US and CT