Performance Comparison of two Machine Learning Methods for Positioning in Analog and Digitally Read-out Slab-detectors

Abstract

The development of Whole and Total-Body Positron Emission Tomography (PET) systems has increased over the last years. These systems aim to increase sensitivity while extending its use to other clinical areas. However, building such systems requires using large volumes of materials and complex signal multiplexing and data analysis processes.In this work, we focus on comparing two PET detector designs that may be part of the potential solution for constructing cost-effective clinical PET systems. Both detectors are based on semi-monolithic crystals and have been developed i3M in Valencia (analog SiPM read-out) and at PMI in Aachen, (digital SiPM read-out). We performed a direct comparison between them by implementing two calibration set-ups and two positioning methodologies based on Neural Networks (NN) and Gradient Tree Boosting (GTB).We report for each set-up the Mean Absolute Error (MAE), bias vector and FWHM of the Point Spread Function. Our results demonstrate that both detectors and positioning algorithms are highly comparable and, thus, potential candidates for being implemented in Whole- and Total-Body PET designs.

Authors 5

1. Florian Mueller Aachen Institute for Experimental Molecular Imaging Department of Physics of Molecular Imaging Systems

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University,Institute for Experimental Molecular Imaging,Department of Physics of Molecular Imaging Systems,Aachen,North Rhine-Westphalia,Germany

   Department of Physics of Molecular Imaging Systems, Institute for Experimental Molecular Imaging, RWTH Aachen University, Aachen, North Rhine-Westphalia, Germany

2. Andrea González-Montoro

   Consejo Superior de Investigaciones Científicas · Instituto de Instrumentación para Imagen Molecular · Universitat Politècnica de València

   Affiliation as printed

   Institute for Instrumentarion in Molecular Imaging (i3M),CSIC-UPV, Detectors for Molecular Imaging,Valencia,Spain

   CSIC-UPV, Detectors for Molecular Imaging, Institute for Instrumentarion in Molecular Imaging (i3M), Valencia, Spain

3. Stephan Naunheim Aachen Institute for Experimental Molecular Imaging Department of Physics of Molecular Imaging Systems

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University,Institute for Experimental Molecular Imaging,Department of Physics of Molecular Imaging Systems,Aachen,North Rhine-Westphalia,Germany

   Department of Physics of Molecular Imaging Systems, Institute for Experimental Molecular Imaging, RWTH Aachen University, Aachen, North Rhine-Westphalia, Germany

4. Volkmar Schulz Aachen Institute for Experimental Molecular Imaging Department of Physics of Molecular Imaging Systems

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University,Institute for Experimental Molecular Imaging,Department of Physics of Molecular Imaging Systems,Aachen,North Rhine-Westphalia,Germany

   Department of Physics of Molecular Imaging Systems, Institute for Experimental Molecular Imaging, RWTH Aachen University, Aachen, North Rhine-Westphalia, Germany

   Hyperion Hybrid Imaging Systems, Aachen, North Rhine-Westphalia, Germany

5. Antonio J. González

   Consejo Superior de Investigaciones Científicas · Instituto de Instrumentación para Imagen Molecular · Universitat Politècnica de València

   Affiliation as printed

   Institute for Instrumentarion in Molecular Imaging (i3M),CSIC-UPV, Detectors for Molecular Imaging,Valencia,Spain

   CSIC-UPV, Detectors for Molecular Imaging, Institute for Instrumentarion in Molecular Imaging (i3M), Valencia, Spain