IEEE 2021 NSS MIC

Please note! All times in the online program are given in New York - America (GMT -04:00) times!

New York - America ()
Jan 26, 2022, 2:55:04 PM
Your time ()
n/a
Tokyo - Asia ()
Jan 27, 2022, 4:55:04 AM
Our exhibitors and sponsors – click on name to visit booth:

To search for a specific ID please enter the hash sign followed by the ID number (e.g. #123).

MIC Mini-Oral IV: Imaging in Therapy / Image Processing

Session chair: Watabe , Hiroshi (Cyric, Tohoku University, Sendai, Japan); Safavi-Naeini , Mitra (Australia's Nuclear Science and Technology Organisation, Australia)
 
Shortcut: MO-04
Date: Wednesday, 20 October, 2021, 11:40 AM - 2:00 PM
Room: MIC - 4
Session type: MIC Session

Contents

Click on an contribution to preview the abstract content.

11:40 AM MO-04-01

BENEdiCTE (Boron Enhanced NEutron CapTurE) Gamma-Ray Detection Module (#913)

A. Caracciolo1, L. Buonanno1, 3, I. D'Adda1, 3, D. Di Vita1, 3, A. Chacon2, M. Kielly2, M. Carminati1, 3, M. Safavi-Naeini2, C. Fiorini1, 3

1 Politecnico di Milano, DEIB, Milano, Italy
2 Australian Nuclear Science and Technology Organisation (ANSTO), Sydeny, Australia
3 Istituto Nazionale di Fisica Nucleare, Sezione di Milano, Milano, Italy

Abstract

We present a gamma-ray detection module for Neutron Capture Enhanced Particle Therapy (NCEPT). The system has been optimised for boron-10 neutron capture agents that can be used for dose enhancement in proton and heavy ion therapy. The goal of the module is to distinguish the photopeak at 478keV from the prompt-gamma emission resulting from the ion-target nuclear interactions. The module consists of a compact 64-channel module, with a large array of SiPM coupled to a 2”×2” cylindrical LaBr3:Ce scintillator crystal (63ph/keV conversion efficiency, 16ns decay time). The electronic front- end ASIC features low-noise processing of photodetector signals, while the pixellated SiPMs detector and individual readout allows for position sensitivity in the crystal. We have characterised the energy resolution of the system experimentally, demonstrating a state-of-the-art energy resolution (3.27% at 662 keV), together with the capability of the FPGA-based DAQ integrated in the module to deploy an external synchronization signal to the ion beam bunches in order to generate anti-coincidence windows. This feature provides a mechanism to distinguish and reject scintillation events from prompt gammas, enhancing the signal-to-background ratio of the spectrometer.

Keywords: Gamma-ray spectroscopy, Boron Neutron Capture Therapy, Silicon Photomultipliers, Ion therapy
11:50 AM MO-04-02

Quantitative helium-beam radiography of an anthropomorphic head and neck phantom exclusively based on silicon pixel detectors (#1068)

M. Metzner1, 2, F. Kehrein1, 2, C. M. Knobloch1, 2, G. Echner1, A. Runz1, B. Ackermann4, 2, S. Brons4, 2, O. Jäkel1, 4, M. Martišíková1, 2, T. Gehrke3, 1

1 German Cancer Research Center (DKFZ), Department of Medical Physics in Radiation Oncology, Heidelberg, Baden-Württemberg, Germany
2 National Center for Radiation Research in Oncology (NCRO), Heidelberg Institute for Radiation Oncology (HIRO), Heidelberg, Baden-Württemberg, Germany
3 University Hospital Heidelberg, Department of Radiation Oncology, Heidelberg, Baden-Württemberg, Germany
4 Heidelberg Ion-Beam Therapy Center (HIT), Heidelberg, Baden-Württemberg, Germany

Abstract

Using quantitative ion-beam radiography to verify treatment plans is a powerful way to reduce uncertainties in ion-beam radiotherapy. In contrast to an X-ray computed tomography scan or an X-ray projection, an ion-beam radiograph shows the line integral of the relative stopping power (RSP) of the tissue directly, which renders a more accurate calculation of the position of the Bragg peak inside of the patient feasible.
We present a quantitative helium-beam radiograph of an anthropomorphic CIRS 731-HN phantom acquired at Heidelberg Ion-beam Therapy center using helium ions with several different energies up to 197.01 MeV/u. The detection system employing the Timepix technology developed at CERN consists of six fully pixelated silicon detectors which are used to measure position and direction of single helium ions in front and back of the imaged object, their time of arrival as well as their energy deposition.
Calibration curves were generated first by measuring the energy deposition of ions which traversed plastic phantoms of well known water equivalent thickness (WET). These curves were then applied to a 48 mm x 24 mm radiograph of the CIRS 731-HN anthropomorphic head phantom, translating the energy deposition values to WET, which is the integral of the RSP along the beam direction.
Comparing the quantitative helium-beam radiograph to a digitally reconstructed radiograph, it becomes apparent that the imaging modality presented in this article is a promising technique offering sufficient spatial resolution and the possibility to measure the integrated relative stopping power directly.

AcknowledgmentThis work is partially funded by the German research council (DFG) under the contract no. JA 1687/11-1
Keywords: helium-beam radiography, anthropomorphic phantom, ion-beam imaging, silicon pixel detectors, particle therapy
12:00 PM MO-04-03

Upstream MLC leaf position detection in complex radiotherapy fields (#645)

J. Pritchard1, J. Velthuis1, 3, L. Beck1, Y. Li1, C. de Sio1, R. Hugtenburg2, 1

1 University of Bristol, Physics, Bristol, United Kingdom
2 Swansea University, Medical School, Swansea, United Kingdom
3 University of South China, Nuclear Science and Technology, Hengyang, China

Abstract

Multileaf collimators (MLC) are an integral component in modern radiotherapy as they dynamically shape the MV photon treatment field and therefore need to be closely monitored to ensure correct treatment delivery. Currently, MLC leaves are calibrated to ±1 mm every 3 months, however leaves can drift beyond this during calibration dates and treatment verification only occurs post-treatment. MAPS detectors are radiation hard for photon and electron irradiation, have high readout speeds and low attenuation which makes them ideal upstream radiation detectors. Previously, we reported on leaf position reconstruction for single leaves using the Lassena, a 12x14 cm2, three side buttable MAPS suitable for clinical deployment. Sobel filter-based methods were used for edge reconstruction. It was shown that correspondence between reconstructed and set leaf position was excellent and resolutions ranged between 60.6±8 and 109±12 mm for a single central leaf with leaf extensions ranging from 1 to 35 mm using 0.3 sec of treatment beam time. Here, we report on leaf edge reconstruction using Sobel filter-based methods in complex leaf configurations, as in clinical use with extensions ranging up to 120 mm. The Lassena detector was placed in the treatment field of an Elekta Agility LINAC with MLC leaves of width 0.5 cm extended into the field creating various leaf configurations. Results show that leaf positions can be reconstructed with resolutions between 78±7 and 149±14 mm at the iso-centre using 0.15 sec long treatment segments. These resolutions significantly exceed current calibration standards.

Keywords: MLC leaf position reconstruction, radiotherapy, MAPS
12:10 PM MO-04-04

Dictionary based MLEM & Simulated Annealing (MSA) GPU-Algorithm for real time PET proton range verification. (#820)

V. Valladolid Onecha1, P. Galve Lahoz1, F. Arias Valcayo1, C. Freijo Escudero1, D. Sanchez-Parcerisa1, 2, P. Ibáñez1, 2, S. España1, 2, L. M. Fraile1, 2, J. M. Udías1, 2

1 Complutense University of Madrid, Nuclear Physics Group & IPARCOS, Madrid, Spain
2 Health Research Institute of the Hospital Clínico San Carlos (IdSSC), Madrid, Spain

Abstract

The full potential of proton therapy will not be realized until a complete control of the uncertainties involved in the determination of the Bragg peak position is mastered.  In-vivo dose verification techniques are being pursued to tackle this problem. Positron Emission Tomography (PET) and Prompt-Gamma Imaging use nuclear activation produced by protons in the patient to obtain dose maps in the irradiated body. The proposed method consists in generating a Monte Carlo data base (Dictionary) with the deposited dose that would be produced by every pencil beam in the treatment plan, as well as the activation image which will be recorded in a PET scanner. Once the Dictionary is calculated a specific ML-EM+SimulatedAnnealing (MSA) algorithm looks for the linear combination of PET activation images of the Dictionary that best fits the observed activation for the plan. This same linear combination is also applied to the pre-calculated doses generating the final dose deposition. The ML-EM is chosen for reconstruction until a consistent solution, combined with Simulated Annealing relaxation step is used in between MLEM iterations to avoid local minima. A GPU implementation allow us to reconstruct the dose from activation data with submillimeter precision in a few seconds. A proton range verification study was conducted using simulated PET and PG images over real proton treatments plan to test our method. PET activity images associated to 1 Gy irradiation provided a dose range localization with accuracy better than 1 millimeter.

Keywords: Proton Therapy, Range Verification, Dose Reconstruction, MLEM algorithm., Simulated Annealing
12:20 PM MO-04-05

Carbon-Ion Beam Monitoring in Depth Using Tracking of Prompt Secondary Ions (#1316)

L. Ghesquière-Diérickx1, 2, R. Félix-Bautista1, 3, A. Schlechter1, 3, T. Gehrke1, 4, L. Kelleter1, 4, P. Soukup7, M. Winter5, 4, K. Herfarth6, 4, J. Debus6, 4, O. Jäkel1, 5, M. Martišíková1, 4

1 German Cancer Research Center (DKFZ), Department of Medical Physics in Radiation Oncology, Heidelberg, Germany
2 University of Heidelberg, Heidelberg Medical Faculty, Heidelberg, Baden-Württemberg, Germany
3 University of Heidelberg, Faculty of Physics and Astronomy, Heidelberg, Germany
4 National Center for Radiation Research in Oncology (NCRO), Heidelberg Institute for Radiation Oncology (HIRO), Heidelberg, Germany
5 Heidelberg University Hospital - Department of Radiation Oncology, Heidelberg Ion-Beam Therapy Center (HIT), Heidelberg, Germany
6 Heidelberg University Hospital, Department of Radiation Oncology, Heidelberg, Germany
7 Advacam s.r.o, Prague, Czech Republic

Abstract

The benefit of the highly precise radiotherapy using carbon-ions can be compromised by even minor changes in the patient geometry such as the internal anatomy or the patient positioning. Therefore, in-vivo monitoring methods of carbon-ion radiotherapy are of great importance to detect and localize such changes which could negatively impair the dose distribution in the patient and thus the outcome of the treatment.

In this contribution, we focus on a contactless, completely non-invasive monitoring method based on the detection and tracking of charged nuclear fragments emerging from irradiated patients. To evaluate the potential of the method, typical carbon-ion treatment irradiations were performed at the Heidelberg Ion-Beam Therapy Center (HIT, Germany) using either anthropomorphic or homogeneous head models. Produced charged fragments were tracked behind the head using a mini-tracker made of two Timepix3 detectors.

To mimic a potential anatomical change in the patient, 2-mm-thick air cavities were introduced in the head model. By tracking the nuclear fragments, significant differences were found in their distribution when introducing the air cavities. These cavities were successfully detected and localized with a precision of 5 mm in depth. Additionally, the developed monitoring method was brought to a status applicable during clinical treatments and it was successfully tested in a real patient irradiation at the HIT facility. The presented results demonstrate the high potential of the method for detecting changes in the patient which can impair the quality of the treatment.

Keywords: particle tracking, prompt secondary ions, carbon-ion beam radiotherapy, hadron therapy, treatment monitoring
12:30 PM MO-04-06

Characterisation of a double-sized Timepix3 mini-tracker in a mixed nuclear fragment field for real-time treatment monitoring in carbon-ion radiotherapy (#1302)

M. Subramanian1, 2, L. Kelleter1, L. Ghesquiere-Dierickx1, 3, T. Gehrke1, 4, R. Felix-Bautista1, 5, G. Echner1, J. Jakubek6, M. Jakubek6, L. Marek6, 7, P. Soukup6, D. Turecek6, M. Martisikova1

1 German Cancer Research Centre DKFZ, Department of Medical Physics in Radiation Oncology, Heidelberg, Baden-Württemberg, Germany
2 Otto-von-Guericke-Universitaet Magdeburg, Faculty of Electro Engineering and Information Technology, Magdeburg, Saxony-Anhalt, Germany
3 Heidelberg University, Medical Faculty, Heidelberg, Baden-Württemberg, Germany
4 Heidelberg University Hospital, Department of Radiation Oncology, Heidelberg, Germany
5 Heidelberg University, Department of Physics and Astronomy, Heidelberg, Germany
6 Advacam s.r.o, Prague, Czech Republic
7 Charles University, Institute of Particle and Nuclear Physics, Faculty of Mathematics and Physics, Prague, Czech Republic

Abstract

Charged nuclear fragments produced during carbon-ion radiotherapy have been proposed to potentially enable so far unachievable real-time treatment monitoring. Observing changes in the fragment track distribution could allow changes in the patient anatomy to be detected during radiotherapy treatment. For this purpose, a novel mini-tracker made of two double-sized Timepix3 hybrid semiconductor pixel detectors has been characterised in a typical clinical mixed radiation field. The allocation of four chips on a single readout interface together with its nanosecond time resolution, as well as the dead-time and noise-free readout allow the reconstruction of individual fragment tracks. The results of the characterisation help to improve the design of a detection system that will be used in an upcoming clinical trial.

Keywords: beam range monitoring, carbon-ion radiotherapy, hybrid semiconductor pixel detector, nuclear fragments, Timepix3
12:40 PM MO-04-07

Optical imaging of proton mini-beams (#199)

S. Yamamoto1, T. Yabe1, T. Akagi2

1 Nagoya University, Nagoya, Japan
2 Hyogo Ion Beam Medical Center, Tatsuno, Japan

Abstract

Proton therapy using mini-beams is a promising method to reduce radiation damage to normal tissue. However, distribution measurements of mini-beams are difficult due to their small structures. Since optical imaging is a possible method to measure high-resolution 2-dimensional dose distribution, we conducted optical imaging of an acrylic block during the irradiation of mini-beams of protons. Mini-beams were made from a proton pencil beam irradiated to 1-mm slits made of tungsten plate. During irradiation of the mini-beams to the acrylic block, we measured the luminescence of the acrylic block using a charge-coupled device (CCD) camera. With the measurements, we could obtain slit beam images that have slit shapes in the shallow area while they were uniform in their Bragg peaks, which was similar to the case of simulated optical images by Monte Carlo simulations. We confirmed that high resolution optical imaging of mini-beams is possible and provides a promising method for efficient quality assessment (QA) of mini-beams as well as research on mini-beam therapy.

Keywords: mini-beam, optical imaging, proton, CCD camera, acrylic block
12:50 PM MO-04-08

Development of a simultaneous imaging system to measure the optical and gamma ray images of Ir-192 source for high-dose-rate brachytherapy (#258)

J. Nagata1, S. Yamamoto1, Y. Noguchi2, T. Nakaya2, K. Okudaira2, K. Kamada3, A. Yoshikawa3

1 Nagoya University Graduate School of Medicine, Department of Integrated Health Science, Nagoya, Japan
2 Nagoya University Hospital, Department of Radiological Technology, Nagoya, Japan
3 Tohoku University, New Industry Creation Hatchery Center, Sendai, Japan

Abstract

In high-dose-rate (HDR) brachytherapy, verification of the Ir-192 source’s position during treatment is needed because such a source is extremely radioactive. One of the methods used to measure the source position is based on imaging the gamma rays from the source, but the absolute position in a patient cannot be confirmed. To confirm the absolute position, it is necessary to acquire an optical image in addition to the gamma ray image at the same time as well as the same position. To simultaneously image the gamma ray and optical images, we developed an imaging system composed of a low-sensitivity, high-resolution gamma camera integrated with a CMOS camera. The gamma camera has a 1-mm-thick cerium-doped yttrium aluminum perovskite (YA1O3: YAP(Ce)) scintillator plate optically coupled to a position-sensitive photomultiplier (PSPMT), and a 0.1-mm-diameter pinhole collimator was mounted in front of the camera to improve spatial resolution and reduce sensitivity. We employed the concept of a periscope by placing two mirrors tilted at 45 degrees facing each other in front of the gamma camera to image the same field of view (FOV) for the gamma camera and the CMOS camera. The spatial resolution of the imaging system without the mirrors at 100 mm from the Ir-192 source was 3.2 mm FWHM, and the sensitivity was 0.28 cps/MBq. There was almost no performance degradation observed when the mirrors were positioned in front of the gamma camera. The developed system could measure the Ir-192 source positions in optical and gamma ray images. We conclude that the developed imaging system has the potential to measure the absolute position of an Ir-192 source in real-time clinical measurements.

Keywords: CMOS camera, gamma ray, HDR brachytherapy, Ir-192 source, scintillation camera
1:00 PM MO-04-09

An Adaptive Deadtime Model for Better Quantification in PET (#860)

M. Aykac1, V. Y. Panin1

1 Siemens Medical Solution USA Inc., Molecular Imaging, Knoxville, Tennessee, United States of America

Abstract

It is essential to have a proper deadtime correction method prior to scatter correction and the reconstruction process. Imperfections in the deadtime model might potentially affect the scatter correction and hence provide inaccurate quantification in the reconstructed images. Even though deadtime is ideally a function of singles rates, correction methods for count losses in PET systems might become more complicated mainly due to processing detector signals especially for piled-up events, changes in scatter distribution for detector pairs at different energy levels and multiplexing the detector readout to minimize the number of electronics channels both in singles and coincidence event processors. Axially long field-of-view PET scanners introduce an additional complexity due to increased sensitivity in detection of scatter.

In this investigation, a new deadtime correction model based on singles rates is discussed with an additional adaptive coefficient. The new model has the flexibility of using two energy windows to accommodate the change in scatter and piled-up scatter events due to various patient/phantom sizes and activity distribution as well. Proposed deadtime model was tested on three phantoms with various shapes and sizes. For those phantoms, accuracy within 2% was achieved up to 150kcps average energy qualified block rate for Siemens Vision Quadra PET/CT scanner.

Keywords: PET, accuracy, deadtime
1:10 PM MO-04-10

Deadtime Correction Method Using Random Coincidences in Block-Pairs for the Siemens 3T MR/BrainPET Scanner (#864)

A. S. M. Issa1, 2, L. Tellmann1, J. Scheins1, A. L. - Montes3, J. L. Herraiz3, N. J. Shah1, 2, C. Lerche1

1 Forschungszentrum Jülich, Institute of Neuroscience and Medicine 4, INM-4, Jülich, North Rhine-Westphalia, Germany
2 RWTH Aachen University, Department of Neurology, Aachen, North Rhine-Westphalia, Germany
3 Complutense University of Madrid, Nuclear Physics Group and IPARCOS, Madrid, Spain

Abstract

Deadtime correction is important to achieve quantitative PET images. The Siemens 3T MR/BrainPET scanner at Forschungszentrum Jülich currently uses a method averaging the dead time over all block detectors and thus ignoring the spatial dependencies and differences in the count rates of the individual blocks. This results in approximative dead time correction with compromising accuracy. In this context, we propose a more accurate dead time correction based on the observation that the block-wise deadtime can be accurately estimated from the detected random coincidence events (using the delayed window technique), which is a recently modified method, however,  in this work we applied the method to the counts per blockpair instead of the counts per ring as in the original work, and it was used for a dedicated head scanner instead of a whole-body scanner. Furthermore, the dead time correction method was corrected for triple coincidences leading to a significant improvement in the accuracy of the method. The maximum deviation between deadtime corrected measured trues and ideally expected trues was only 1.5 % in the worst case. Without triple correction, the maximum deviation between measured trues and expected trues was 4 %.

Keywords: deadtime, quantitative images, 3T MR/BrainPET scanner, triple coincidences, random coincidence events
1:20 PM MO-04-11

Correcting spatial positioning errors in pre-defined hardware attenuation correction maps for PET/MR (#894)

P. J. Schleyer1

1 Siemens Medical Solutions USA, Inc., Knoxville, Tennessee, United States of America

Abstract

A spatially accurate attenuation correction (AC) map of materials that all acquired PET lines of response (LORs) traverse is required for attenuation correction. Materials include the patient, patient table, and for PET/MR systems local RF coils which are used that cannot be neglected due to their design. Patient attenuation coefficients are usually derived from MR images, however the remaining non-radioactive hardware is typically corrected by referring to pre-defined CT-based AC maps. A spatial offset between the true and assumed position, for example a vertical table displacement resulting from patient weight, can produce attenuation correction errors.

A simple approach that transforms the pre-defined hardware AC map to its true position is presented. Using the principle of consistency criteria, the spatial transform that minimizes azimuthal variations in attenuation corrected LORs is determined. Only LORs which are localized to specific coordinates are considered to reduce computational demand.

A phantom experiment simulated vertical displacement of the patient table and MR head-neck coil by introducing offsets ranging from -16.5 mm to 16.5 mm (in 1 pixel increments) to the hardware AC map. Attenuation correcting with offset hardware AC maps produced structured PET artefacts with up to 6.4 % error. The proposed method accurately corrected the offset, eliminating the error in all cases.

Keywords: PET/MR, attenuation correction
1:30 PM MO-04-12

Statistical CT sinogram generation from time-of-flight PET data using kernel methods in the projection space (#1175)

Y. Zhu1, G. Wang1

1 University of California, Davis, Department of Radiology, Sacramento, California, United States of America

Abstract

Accurate attenuation correction plays an essential role for quantitative PET imaging. For PET/CT system, attenuation correction is usually performed by converting x-ray CT images for 511 keV gamma-ray photons. However, such conversion may introduce potential errors, e.g. in contrast-enhanced PET/CT and bone imaging. The maximum likelihood attenuation correction factor (MLACF) method has shown a potential to generate improved reconstruction results by directly estimating a “gamma-ray CT” attenuation sinogram from time-of-flight PET emission data. However, MLACF suffers from data noise. In this work, we propose to use kernel methods to improve MLACF for gamma-ray CT sinogram generation by incorporating the x-ray CT sinogram as prior information. Different from the existing kernel methods for PET activity image reconstruction that are all operated in the image space, the kernel method for MLACF is in the projection space. Both the popular radial Gaussian kernel and a highly constrained back-projection (HYPR) kernel were used in this work. We conducted computer simulations to evaluate the two kernel MLACF methods and compared them with the conventional MLACF and CT-converted methods. Results indicate kernel MLACF is able to generate better quantitative results compared to other methods.

AcknowledgmentThis work is supported in part by NIH under grant no.  R21EB027346
Keywords: Attenuation correction, kernel method, MLACF, TOF PET
1:40 PM MO-04-13

A novel attention-based convolutional neural network for joint denoising and partial volume correction of low-dose PET images (#378)

M. Azimi1, H. Arabi2, A. Kamali-Asl1, M.R. Ay3, H. Zaidi2

1 Shahid Beheshti University, Department of Medical Radiation Engineering, Tehran, Iran (Islamic Republic of)
2 Geneva University Hospital, Division of Nuclear Medicine & Molecular Imaging, Geneva, Switzerland
3 Tehran University of Medical Sciences, Department of Medical Physics and Biomedical Engineering, Tehran, Iran (Islamic Republic of)

Abstract

To achieve a high-quality PET image for diagnostic purposes, a standard dose of radioactive tracer must be injected into the patient's body, which increases the risk of radiation damage. However, reducing the tracer dose leads to poor quality of the PET images and noise-induced quantitative bias. Another concern for quantitative PET imaging is partial volume effects (PVE) which are a consequence of the inherently limited spatial resolution of PET and introduce large biases especially for structures with size of the same order of point spread function (PSF). In this work, we aim to use deep learning-based methods to propose a jointly approach for predicting full-dose and PVE-corrected (FD+PVC) images from the low-dose (LD) counterparts. Deep learning methods generally use the whole image as input for network learning, while the most common PVC methods such as GTM (geometric transfer matrix) aiming at extracting PVE-corrected activity concentration values for user-defined volumes of interest (VOIs), so the focus is only on the specific areas of the image, and when the whole image is used the training of the net-work may be sub-optimal. In this work, we propose an attention-based convolutional neural network (ATB-Net) to predict full-dose and PVE-corrected (FD+PVC) from low-dose (LD) PET images via focusing the attention of the network on the Automat-ed Anatomical Labeling (AAL) brain regions. By evaluating PSNR and RMSE metrics ATB-Net being 20.91% and 12.39%, respectively better than the U-Net. The region-wise analysis showed that the difference between the relative bias of U-Net and ATB-Net were all statistically significant (p<0.05). In the case of absolute relative bias, more than half of the regions have significant differences between the results (p<0.05). Therefore, the proposed ATB-Net would be able to simultaneously performed correction for the partial volume effects owing to access to the anatomical regions of the brain and for denoising.

Keywords: Low dose PET, attention map, deep convolutional neural network, partial volume correction
1:50 PM MO-04-14

Towards development of image-derived input function from carotid arteries for the NeuroEXPLORER (NX) (#1286)

P. Honhar1, T. Li2, J. Schmall3, R. Carson1, A. Hillmer1, T. Feng3

1 Yale University, New Haven, Connecticut, United States of America
2 University of California, Davis, Davis, California, United States of America
3 United Imaging Healthcare America, Houston, Texas, United States of America

On behalf of the NX Consortium

Abstract

For brain PET scanners with a limited axial field of view, carotid arteries (CA) are the logical choice for acquiring an image-derived input function (IDIF). However, CA quantification has significant partial volume effects, which has limited the use of IDIFs on most PET scanners. The NeuroEXPLORER (NX) is a next generation brain PET scanner which promises an effective 10 x improvement in sensitivity compared to conventional brain PET scanners. We hypothesize that the performance improvements in the NX should allow for accurate quantification of CA derived input function. To simulate NX data, uEXPLORER’s FDG dynamic scan datasets are reconstructed with a limited acceptance angle. Algorithms which were previously developed for CA-based IDIF extraction in uEXPLORER were repurposed for the NX. The carotid arteries were manually identified using the first 60 s of reconstructed images. A thresholded iterative deconvolution approach was used for CA segmentation at each slice. The segmented CA regional time activity is partial volume corrected using a geometric transfer matrix method, without need for a blood sample. IDIF from the descending aorta serves as the gold standard. The results show that partial volume correction significantly improves the area-under-curve (AUC) ratio (CA-IF:Aorta-IF) for both uEXPLORER and NX data. Based on AUC ratios, the accuracy of the CA-IF obtained from NX data is slightly less compared to the CA-IF derived from uEXPLORER data. A 4-dimensional (4D) digital phantom that simulates a dynamic FDG scan was also created for this study. Analytical projections were used to reconstruct the phantom for the NX. The CA location and its separation from the jugular vein was similar between the reconstructed images and ground truths from the phantom.

AcknowledgmentResearch reported in this work was supported by the National Institute of Biomedical Imaging and Bioengineering of the National Institutes of Health under award number U01EB029811.
Keywords: NeuroEXPLORER, Image-derived Input Funtion, Carotid Artery, Partial Volume Correction

Our exhibitors and sponsors – click on name to visit booth: