4-8 July 2016
Kramer Law building
Africa/Johannesburg timezone
Paper Review: Initial screening in progress
Home > Timetable > Session details > Contribution details
PDF | XML | iCal

A genetic algorithm approach to enhancing the performance of a PET detector array

Presented by Prof. Simon CONNELL on 6 Jul 2016 from 16:10 to 18:00
Type: Poster Presentation
Track: Track F - Applied Physics
Board #: F.153

Abstract

The MinPET project aims to locate diamonds within kimberlite by activating carbon within kimberlite, then using Positron Emission Tomography (PET) to image carbon density. Distinguishing small diamonds from the background depends crucially on the accurate reconstruction of detector hit positions. This reconstruction is subject to two kinds of errors: local errors, where the position of a hit within a particular detector is not accurately reconstructed from the incoming photomultiplier tube signals, and global errors, where the internal parameters that describe the physical location and orientation of detector pixels do not accurately match reality. Because of the large number of detectors in a full MinPET unit, there are too many parameters involved to feasibly adjust them by hand. We have therefore developed a custom genetic algorithm that iteratively evolves detector parameters in order to optimise the image quality. Results are presented from before and after the optimisation is performed, indicating that image accuracy and resolution are improved. This algorithm could be employed periodically in an industrial or medical setting to automatically correct for detector movements or calibration drift.

Award

No

Level

N/A

Paper

Yes

Permission

Yes

Place

Location: Kramer Law building
Address: UCT Middle Campus Cape Town
Room:


Primary authors

More

Co-authors

More