XCIST User Survey

This survey was intended for current and potential users of the CatSim CT and X-ray simulation tool. The survey contained ten questions to understand current and future use cases for CatSim/XCIST, and to prioritize future development of its capabilities and features. Responses to these questions will be kept confidential and used for research purposes only. We may use respondent's contact information to follow up for clarification on specific answers, but only if necessary for internal evaluation. However, under no circumstances will individual responses be identifiable in the reported results. All survey findings will be reported/published anonymously.

Survey questions are listed below, for reference:

Your position (multiple choice)

• Undergraduate student
• Graduate student
• Postdoc
• Professor
• R&D scientist
• R&D manager
• Other (Please specify)

How relevant and important are the following CatSim/XCIST use cases for you?

The XCIST users' survey results for this question are shown in the following graph, ranked in descending average score order:

(Respondents had five choices for five topics):

• is essential, My research would not be possible without CatSim/XCIST.
• is important. Some of my research would not be possible without CatSim.
• greatly improves the efficiency/relevance of my research.
• is helpful for my research, but not so essential.
• is not relevant for my research.

1. Development/evaluation of new X-ray and CT systems/methods/algorithms:
2. Deep Learning (DL) training data generation - using data that others have generated
3. Deep Learning (DL) training data generation - performing data generation myself
4. Virtual imaging trials
5. Educational purposes (e.g. teaching CT class)

How useful/impactful would each of the following CatSim/XCIST new features be for you?

The XCIST users' survey results for this question are shown in the following graph, ranked in descending average score order:

(Respondents had five choices for 15 topics):

• is essential. This will be our biggest reason for using CatSim/XCIST.
• is important. This will lead to important additional use of CatSim/XCIST.
• is a great enhancement. This will increase/improve our use of CatSim/XCIST.
• is nice to have, but not critical in our use of CatSim/XCIST.
• is unnecessary. I don’t expect to use it.

1. Photon-counting CT
2. Phase-contrast & darkfield imaging
3. Fully validated dose estimation tools
4. Deep learning reconstruction
5. Image quality evaluation tools (phantoms + metrics)
6. Population of patient datasets (phantoms) and virtual lesions
7. Generative AI for virtual, realistic phantom generation
8. Task-based performance evaluation (Channelized hotelling observer, DL, …)
9. Graphical user interface for intuitive use of CatSim/XCIST
10. Cloud-based CatSim/XCIST deployment (through web interface)
11. Hardware acceleration support such as GPU implementation
12. Interfaces with DL toolkits (e.g. Pytorch) to facilitate DL training
13. Computation of electron density and stopping power for photon and proton therapy planning simulation
14. Realistic CT scanner models across multiple vendors
15. Other (Please specify)

What barriers are preventing you from using CatSim/XCIST, if any? (e.g., software license, coding language)

Summary of the responses are as follows:

Difficulties Lack of user-friendly tutorials and documentation.

• Difficulty in learning to use the code, especially for short-term interns.
• Obscure documentation.
• Time-consuming to get started.
• Challenging to compile the code.

Usability

• Needs a more comprehensive documentation that explains various variables and how they affect the generated data.
• It would be helpful if it were easier to discover capabilities.

Functionality

• Lacks features like helical reconstruction methods.
• Potential gap between the actual CT system and CatSim.

Performance

• The CPU-based version is very slow.

Other answers

• More examples, simulations and training resources.
• Hardware acceleration for faster simulations.
• Porting more features from the proprietary version to the open-source version.

Do you have any other suggestions for CatSim/XCIST?

Summary of the suggestions are as follows:

Usability

• Make CatSim user friendly with minimal coding required.
• Create video tutorials to teach people how to use CatSim.
• Develop a user interface (UI) using AppDesigner from Matlab.
• Provide more detailed descriptions of configuration parameters.
• Create a database of archived simulation setups.

Functionality

• Expand CatSim's capabilities to include simulating tissue/organ motion and angiogenesis/neovacularization related to anthrosclerosis and/or neoplasma.
• Add a GPU-powered monte-carlo capability and embrace large scale distributed computing.
• Develop next generation CT scanner modeling and validate these scanner models with actual experimental data.
• Include tomosythesis recon.
• Add more photon interaction processes.

Documentation

• Improve documentation with more examples and detailed instructions.
• Create a link from the GitHub repository to "marketing material" (publications describing use cases, web page highlighting its use wish pictures, etc.) Performance
• Improve speed by making CatSim more modular and by prioritizing hardware.
• acceleration/parallelization/GPU computational capabilities.

Open Source

• Integrate algorithms developed by various non-computer science engineers/researchers into one code base.

Other

• Run CatSim on a cloud system.
• Develop an interface to large models and foundation models.