Summary

This project is the code used in a presentation made at the CHU de Quebec - Université Laval on using programming as a tool.

Some radiation therapy data is analyzed using a few techniques:

1. Microsoft Excel (TM)
2. Vanilla Python code
3. Numpy Python code
4. C++ code

Data Generation

The data represents a dose profile in radiation therapy, i.e. a plot of dose as a function of lateral position in a medical radiation beam. Ignoring the file header, it consists of two columns representing position (cm) and dose (%).

The original data I had available was not suitable for the presentation, since it had already been analysed and modified for our purposes. In reality, the data to be analyzed has a variable number of entries and the step between the points varies. Of course, real data is not 100% symmetrical, since this is what the analysis is calculating. Thus, I have created suitable data for the presentation. This has been done mostly using python (with a bit of manual editing, hey, I'm not perfect).

In the data_generation folder, you will find two sub-folders and a python script. In the 'orig' subfolder, you can find the original treated data, which would be useless for the purposes of the presentation. In the 'var_lengths' subfolder, you can find the data with a variable number of entries, which was made by hand. I just copied the files and removed a centimeter of data from each end. Could have been done programmatically as well.

The last part of the data generation is the script. This script will take the variable length data and read it. It will then introduce a position shift on every point (selected randomly, but the same for every position) and some noise in the dose for the inner 80% to simulate non symmetrical data. The new data will be written in the data folder of the project.

Analysis

Excel

This Microsoft Excel (TM) spreadsheet is probably not 100% optimized and I'm sure some people will say I could have done better. That being said, it does calculate the correct symmetry for the data. To calculate a new profile, you have to:

1. Copy one of the data sheets. Rename it to something appropriate.
2. Paste your data into the columns with the red background. Some caveats:
3. You must make sure to remove the excess data if some exists and leave only your data.
4. If you change the location of the first cell, you must adjust the appropriate columns in the Analyse sheet.
5. Create a new line in the Analyse sheet for your new data sheet. Adjust the name column.
6. Update every cell in your data sheet which refers to the Analyse sheet to have the appropriate line number. They are with a blue background.

This should then yeild the symmetry in the last used colum of your data sheet.

Programs

Every program needs to be provided at least one (or more) file name. It opens the file as an ASCII file and attempts to read radiation profile data from it. It then calculates the symmetry and reports it to standard output.

Vanilla Python

The program is divided into two files : lib.py and analyze.py. The analyze.py file is the one containing the main program and the lib.py file contains the more generic helper functions. It does not contain a hash/bang symbol, meaning it is not considered an executable and must be run explicitely through Python.

python analyze.py "my file name"

Numpy Python

Approx

The program uses Numpy functions instead of the hand written ones, but implments the same algorithm with the same command line interface.

python analyze.py "my file name"

Gradient

The program uses Numpy functions and a difference algorithm to find the width and center of the profile. The algorithm is based on the maximum and minimum of the derivative of the profile to extract the two 50% positions. Then, the rest of the algorithm is pretty much the same.

python compute_sym.py "my file name"

C++

The program is a single cpp file which contains everything. It relies only on the STL (Standard Template Library), but does use C++14 features so should be compiled with the appropriate compiler options on some platform (Clang and GCC). To compile into an executable:

• GCC : g++ -std=c++1z main.cpp -o main
• Clang : clang++ -std=c++1z main.cpp -o main

The executable can the be run as

main "my file name"