Name: From imaging algorithms to quantum methods Seminar
Start: 2025-09-29T10:00:00+02:00
End: 2025-09-29T11:30:00+02:00
Location: Zoom

Hide

# Title: Proton computed tomography from sandwich time-of-flight data

## Presenter: Aurelien Coussat
## Date: 30.09 2025

## Participants:

Aurelien Coussat (AC)
Wojciech Krzemień (WK)
Konrad Klimaszewski (KK)
Roman Shopa (RS)
Lech Raczyński (LR)
Kamil Dulski (KD)
Oleksandr Fedoruk (OF)

## Questions/Remarks:

WK: Are all your building blocks functions differentiable?
AC: Yes, all are differentiable.

WK: Does the autograd know how to differentiate all functions? Do you provide both the function and its derivative?
It knows in my case, cause the "differentiation" is trivial?
AC: It is exactly how I finally implemented it. I provided both function and derivative as building blocks.

KK: How is the Most Likely Path estimated? Is it purely geometrical?
AC: Yes. It was investigated a lot, and it seems, in some sense, sufficient.

WK: Did you think about some regularisation?
AC: Yes, we considered it, especially for a specific phantom where we have issues with the conversion.

LR: How do you initialise your algorithm?
AC: For a moment, we start with the empty image.

There are minutes attached to this event. Show them.

- 10:00 → 11:00
  
  Proton computed tomography from sandwich time-of-flight data 1h
  
  Proton computed tomography (pCT) is an imaging modality which provides several benefits including the accurate estimation of the relative stopping power (RSP) distribution for proton therapy treatment planning. Conventional proton CT scanners measure the energy loss of each proton individually, along with their positions and directions before and after the object, to estimate the integral of the RSP along each proton’s most likely path (MLP).
  
  However, keeping reasonable acquisition times and moderate manufacturing costs with this design is challenging. As a potential solution, a novel design was recently proposed where the RSP distribution is inferred from the time-of-flight (TOF) of protons between two detectors sandwiching the object, without needing a residual energy tracker. However, a major difficulty stems from this new design: to each measured TOF corresponds infinitely many possible water-equivalent path lengths (i.e., RSP integrals) because the velocity of the proton depends on the order of the materials encountered along the MLP. Consequently, no reconstruction algorithm is currently available for this set-up.
  
  In this talk, I will present two reconstruction procedures that are currently under investigation at the CREATIS laboratory (Lyon, France).
  
  Speaker: Aurélien Coussat
  
  coussat_seminar.pdf
- 11:00 → 11:30
  
  Discussion 30m

From imaging algorithms to quantum methods Seminar

https://cern.zoom.us/j/66151941204?pwd=n7upvvZYibexBhbtyn5kvTpy36L0Wo.1

Zoom