Nonnegative Gaussian process tomography for generalized segmented planar detectors

TL;DR

This paper introduces a nonnegative Gaussian process tomography method for high-resolution 3D image reconstruction in segmented planar detectors, with applications in proton therapy dose imaging.

Contribution

It extends Gaussian process tomography with nonnegative constraints to 3D and generalized detector segmentations, tailored for proton therapy imaging.

Findings

Effective 3D dose reconstruction demonstrated

Fast approximate nonnegative constraint application developed

Applicable to complex detector segmentation and readout mappings

Abstract

The concept of Gaussian process tomography along with nonnegative constraints is applied in the context of high-resolution image reconstruction using segmented planar detectors with few readout channels.Expanding on the concept of 2-D projections onto strip-like readout segmentations, 3-D projections as well as more generalized detector segmentation and readout channel mappings are explored. A focus is placed on reconstructing dose distributions in proton therapy pencil beam scanning, and a fast, approximate approach to applying nonnegative constraints is developed and motivated for use in proton therapy beam imaging.