how can i start in so many repositories

[Little-Stars520]

Hello, I am a novice medical physicist and not familiar with python. How should I start in this so many repositories ? I want to learn how the dose is calculated in current TPS and how MLC is inversely optimized? Thank you for taking time out of your busy schedule to answer

[QuangMacHUST]

Thank you for your question. Currently, the TPS is still in the process of development and refinement. Regarding your two questions about dose calculation and MLC:
Dose calculation: We measure actual dose distribution data from various depths. I have taken data from the Varian TrueBeam and integrated it into the system. From a system developer’s perspective, this is not as straightforward as it might seem from a medical physicist’s viewpoint. We need to simulate and calculate interactions (specifically in dose/physics), convert HU density to electron density for computation purposes, implement dose calculation algorithms (found in the dose/algorithms folder), and address many other related aspects such as contouring, retrieving organ contour data, and other system-related components.
Regarding your second question: Think of MLC optimization as just one part of the broader dose optimization problem. Optimizing a radiotherapy plan across multiple organs is a complex problem with many parameters. You need to define the problem with objectives, optimization techniques, and parameters like MLC thickness, HyperArc, etc. We set these up, and there are equations to solve that have been derived from scientific papers. You can refer to the GitHub repository at github/repo/porty for more details.

[QuangMacHUST]

Where are you from? I come from Vietnam, and I hope this repository is accessible to everyone for educational and research purposes

[Little-Stars520]

Thank you for your question. Currently, the TPS is still in the process of development and refinement. Regarding your two questions about dose calculation and MLC: Dose calculation: We measure actual dose distribution data from various depths. I have taken data from the Varian TrueBeam and integrated it into the system. From a system developer’s perspective, this is not as straightforward as it might seem from a medical physicist’s viewpoint. We need to simulate and calculate interactions (specifically in dose/physics), convert HU density to electron density for computation purposes, implement dose calculation algorithms (found in the dose/algorithms folder), and address many other related aspects such as contouring, retrieving organ contour data, and other system-related components. Regarding your second question: Think of MLC optimization as just one part of the broader dose optimization problem. Optimizing a radiotherapy plan across multiple organs is a complex problem with many parameters. You need to define the problem with objectives, optimization techniques, and parameters like MLC thickness, HyperArc, etc. We set these up, and there are equations to solve that have been derived from scientific papers. You can refer to the GitHub repository at github/repo/porty for more details.

@QuangMacHUST
Thank you for your answer.
Thank you very much for writing this project. I'm very interested in learning this project, and at the moment I only see matrad, but I think your project introduction is more attractive than matrad, with more dose calculation algorithms. I'm very eager to learn from your project
I would like to ask which file of which project should I start with to complete the entire dose calculation process? Because there are too many projects here ( [QuangTPS]; [QuangStationV2];...)and there are a lot of python files for each project. Additionally, I found that the demo_guide of the project is 404, and I was very confused about how to start running a complete radiotherapy example?
I suggest you write a manual on how to run different examples and list relevant documents, so that it is more friendly to newcomers

[Little-Stars520]

Where are you from? I come from Vietnam, and I hope this repository is accessible to everyone for educational and research purposes

I come from Singapore, I think your project will gain a lot of stars

[QuangMacHUST]

Thank you so much for your appreciation. However, unfortunately, I’m a 22-year-old student living in Vietnam. I’m currently studying at Hanoi University of Science and Technology, and I’m in my fourth year. At the moment, I feel that career opportunities in the field of medical physics are very limited, and I’m unable to find a job. I’m planning to switch to biomedical engineering and completely stop working on this project.

Although I really want to study abroad in countries around the world, it seems challenging since I’m on my own and don’t have anyone to guide me. I’m still passionate, but unfortunately, I need to put this project on hold due to the demands of making a living.

If possible, could you share this project with your professors or colleagues for them to review and evaluate? I’m still in the development process, but overall, the project already provides a general overview. However, there are still quite a few bugs because I’ve been updating features, which has made it a bit messy.

It’s really a pity that I have to abandon this project. Thank you, my first reader.

[Little-Stars520]

Thank you so much for your appreciation. However, unfortunately, I’m a 22-year-old student living in Vietnam. I’m currently studying at Hanoi University of Science and Technology, and I’m in my fourth year. At the moment, I feel that career opportunities in the field of medical physics are very limited, and I’m unable to find a job. I’m planning to switch to biomedical engineering and completely stop working on this project.

Although I really want to study abroad in countries around the world, it seems challenging since I’m on my own and don’t have anyone to guide me. I’m still passionate, but unfortunately, I need to put this project on hold due to the demands of making a living.

If possible, could you share this project with your professors or colleagues for them to review and evaluate? I’m still in the development process, but overall, the project already provides a general overview. However, there are still quite a few bugs because I’ve been updating features, which has made it a bit messy.

It’s really a pity that I have to abandon this project. Thank you, my first reader.

@QuangMacHUST
I'm really sorry. In fact, few people are willing to put in the effort to open - source the codes related to traditional photon dose calculation. Currently, most professors focus on content with academic value. So, it's very difficult for me to find codes on how to implement traditional photon dose calculation. Could you offer some assistance with such codes? For example, those related to pencil beam algorithms, CCC algorithms, etc. I don't know how to reproduce these algorithms just by reading the papers. Moreover, I'm also quite confused about how to incorporate the different radiation fields formed by sliding MLC into the dose calculation.

[QuangMacHUST]

1999-Ahnesjo_DoseCalculations.pdf
aaa algorithm.pdf
ccc-2013.pdf
EclipseAlgorithms13.6_RefGuide.pdf
method-calculate-dose-external-beam.pdf
overview.pdf
The only way is to read and read more to understand all the formulas and algorithms. I know you’re confused by the extremely long and complex algorithmic formulas, but this is advanced mathematics—it’s not that difficult. Just break it down slowly and understand each component one by one. I know it’s challenging, but you have to read and reflect on it. Start with the simpler algorithms first, beginning with the physical interactions using Type A algorithms in this video: https://www.youtube.com/watch?v=kNT8B-SvCMs&t=302s.

[Little-Stars520]

1999-Ahnesjo_DoseCalculations.pdf aaa algorithm.pdf ccc-2013.pdf EclipseAlgorithms13.6_RefGuide.pdf method-calculate-dose-external-beam.pdf overview.pdf The only way is to read and read more to understand all the formulas and algorithms. I know you’re confused by the extremely long and complex algorithmic formulas, but this is advanced mathematics—it’s not that difficult. Just break it down slowly and understand each component one by one. I know it’s challenging, but you have to read and reflect on it. Start with the simpler algorithms first, beginning with the physical interactions using Type A algorithms in this video: https://www.youtube.com/watch?v=kNT8B-SvCMs&t=302s.

@QuangMacHUST
In fact, I've read many textbooks. They basically only mention the explicit Terma and then perform convolution and superposition at each point. This method is very easy to understand. However, when I look at current TPS, such as Eclipse, the Varian accelerator only collects data at certain depths and for square fields of certain sizes during machine commissioning to complete patient dose calculations. Moreover, I don't know how to account for the changes in the field size and transmission caused by the movement of the MLC. Also, the MLC only considers the size at the isocenter. May I ask which codes and literature you referred to during your learning?

In addition, would you mind giving me your contact information for further communication?

[Little-Stars520]

1999-Ahnesjo_DoseCalculations.pdf aaa algorithm.pdf ccc-2013.pdf EclipseAlgorithms13.6_RefGuide.pdf method-calculate-dose-external-beam.pdf overview.pdf The only way is to read and read more to understand all the formulas and algorithms. I know you’re confused by the extremely long and complex algorithmic formulas, but this is advanced mathematics—it’s not that difficult. Just break it down slowly and understand each component one by one. I know it’s challenging, but you have to read and reflect on it. Start with the simpler algorithms first, beginning with the physical interactions using Type A algorithms in this video: https://www.youtube.com/watch?v=kNT8B-SvCMs&t=302s.

In fact, I think reading code is the fastest way to learn. If you have more code to share, could you please share it with me?

[QuangMacHUST]

1999-Ahnesjo_DoseCalculations.pdf aaa algorithm.pdf ccc-2013.pdf EclipseAlgorithms13.6_RefGuide.pdf method-calculate-dose-external-beam.pdf overview.pdf The only way is to read and read more to understand all the formulas and algorithms. I know you’re confused by the extremely long and complex algorithmic formulas, but this is advanced mathematics—it’s not that difficult. Just break it down slowly and understand each component one by one. I know it’s challenging, but you have to read and reflect on it. Start with the simpler algorithms first, beginning with the physical interactions using Type A algorithms in this video: https://www.youtube.com/watch?v=kNT8B-SvCMs&t=302s.

@QuangMacHUST In fact, I've read many textbooks. They basically only mention the explicit Terma and then perform convolution and superposition at each point. This method is very easy to understand. However, when I look at current TPS, such as Eclipse, the Varian accelerator only collects data at certain depths and for square fields of certain sizes during machine commissioning to complete patient dose calculations. Moreover, I don't know how to account for the changes in the field size and transmission caused by the movement of the MLC. Also, the MLC only considers the size at the isocenter. May I ask which codes and literature you referred to during your learning?

In addition, would you mind giving me your contact information for further communication?

You're correct that during beam commissioning, TPSs like Eclipse typically use measured data (PDDs, profiles, output factors) for square fields of specific sizes at certain depths. From there, the system builds a dose model by fitting the data to a mathematical kernel (e.g., pencil beam, collapsed cone, or Monte Carlo-based).

When the field size or shape changes—especially with MLCs—the TPS doesn’t just interpolate blindly. It uses:

Fluence modeling: MLC transmission, leaf scatter, tongue-and-groove effects, and interleaf leakage are all modeled using separate datasets.

Effective field size corrections: Algorithms adjust output factors using methods like equivalent square field or model-based scaling.

Off-axis correction: Even though MLC defines shape at the isocenter, the model accounts for divergence and off-axis variations (e.g., via beamlet-based modeling).

Beamlet-based dose computation (in algorithms like AAA or Acuros): Each beamlet has its own kernel, modulated by MLC transmission and shape.

You can imagine the MLC as a moving block. When techniques like IMRT, VMAT, or SRS are performed, the MLC shapes the field according to our requirements, and the machine delivers the beam with a dose rate in MU per second corresponding to that field. This technique is often referred to as ‘stop-shot-move’.

Regarding aspects such as radiation leakage, maximum field size, number of leaves, leaf thickness, transmission through the leaves, and the tongue-and-groove effect of the MLC—these parameters may or may not be publicly available, depending on the vendor. For example, Varian has different types of MLCs like the 80-MLC, 120-MLC, HD-MLC, and Elekta has the 160-MLC. Whether or not documentation is publicly available, I’m not sure, since I have a Varian account and can download those materials.

As for the question of whether changes in MLC affect dose—yes, they do. But we have various methods to address this. For instance, I divide the patient’s irradiated volume into voxels and calculate the dose received by each voxel based on the influence of the field shape and MLC modulation.

To calculate the dose influenced by the MLC shaping the field over the patient, you need to measure data that simulate transmission through the leaves, scattering at the leaf tip, and interleaf leakage. Only when calculating fluence from techniques like IMRT or VMAT do you actually need to perform such complex modeling.

[QuangMacHUST]

1999-Ahnesjo_DoseCalculations.pdf aaa algorithm.pdf ccc-2013.pdf EclipseAlgorithms13.6_RefGuide.pdf method-calculate-dose-external-beam.pdf overview.pdf The only way is to read and read more to understand all the formulas and algorithms. I know you’re confused by the extremely long and complex algorithmic formulas, but this is advanced mathematics—it’s not that difficult. Just break it down slowly and understand each component one by one. I know it’s challenging, but you have to read and reflect on it. Start with the simpler algorithms first, beginning with the physical interactions using Type A algorithms in this video: https://www.youtube.com/watch?v=kNT8B-SvCMs&t=302s.

In fact, I think reading code is the fastest way to learn. If you have more code to share, could you please share it with me?

i don't have reference code

[QuangMacHUST]

@Little-Stars520 were you satisfied with my answer? I don't have teacher in this major

[Little-Stars520]

@Little-Stars520 were you satisfied with my answer? I don't have teacher in this major

@QuangMacHUST
Thank you for your answer. In fact, without the code, I think these are really difficult for me to understand. I only understand that if I know the Terma and kernel values of each pixel, I can calculate the dose through convolution and superposition, which I think is the simplest method. However, for most TPS, such as Eclipse. In fact, I have read the AAA documentation of Varian, but even after reading it, I still don't know how it is implemented. That's why I'm looking for code on GitHub to study.