Quantification of positron emission tomography brain imaging using 18F-Fallypride: a simulation

Mohlapholi, Mohlapoli Stadium (2019-04)

Thesis (MSc)--Stellenbosch University, 2019.

Thesis

ENGLISH ABSTRACT: 18F-fallypride is of interest in neurological and psychiatric diseases to visualise dopamine D2/D3 receptors in striatal and extrastriatal regions. Several simplified methods of quantifying positron emission tomography (PET) brain imaging using 18F-fallypride have been described and used on real data using non-invasive methods, but a systematic analysis of their effect on quantification has not been performed. In this study, mathematical simulations were used to study this effect on quantification using different models to quantify 18F-fallypride PET in the human brain. The specific uptake areas of interest used in this study for a human brain are the putamen (high receptor density region), thalamus (moderate receptor density region) and temporal cortex (low receptor density region). Materials and methods All simulations were performed using Matlab (version R2013a (8.1.0.604) win 64-bit software; MathWorks, Inc.). Simulations of realistic measurements were performed by modelling varying frame duration, decay of the tracer, and varying noise levels, starting from ideal tissue curves. A modelled input function was used for the generation of ideal tissue curves. Quantification was carried out with arterial blood sampling models: a 2T4k (two-tissue (or three-compartment) reversible model was used and a Logan graphical analysis was applied. Using the cerebellum as the reference region, we studied the following reference region models: Logan graphical analysis (Loganref) using a reference region, the reference tissue model (RTM) and the simplified reference tissue model (SRTM). Finally, the standard uptake value ratio (SUVR) was studied as the simplest method. For the assessment, the results were analysed using a correlation analysis and a Bland-Altman analysis and the relative error (%) and the 5th and 95th percentiles in high/moderate/low receptor density region of the brain were determined. Validation of each method was done in terms of bias and variance. Results The study showed that in all cases the ground truth method and the graphical method using an arterial input function were nearly identical. The overall accuracy and variability of various methods were determined successfully. Loganref was the most accurate, with the highest precision as the replacement of the invasive arterial blood sampling in low/moderate receptor density regions. While SRTM gave high precision in high receptor density region, SUVR calculations produced relatively large errors in all receptor density regions used in this study. The effect of SNR (signal to noise ratio) on quantification was clearly observed since the bias in all reference region methods increased with increasing noise. When the noise level is too high, bias may become too large. Investigating the effect of the models’ performance under a continuous sampling scheme showed the stability in the kinetics of 18F-fallypride using non-invasive reference region methods when the scan length (continuous sampling scheme) was > 150 min and > 120 min in high and in low/moderate receptor density regions respectively. To achieve a good compromise between accuracy and patient comfort, breaks can be introduced at pre-determined intervals. Best results were achieved when the total scan time was 90 min and the total break time was 150 min (i.e. 30 min scan + 30 min break + 30 min scan + 120 min break + 30 min scan). Conclusion Simplified models can be used to provide useful estimates of dopamine transporters that are comparable to methods using arterial blood sampling. However, these models should be used with great care as they can be affected by the noise in the data, the length of the scan duration and the length and position of the breaks in the imaging sequence.

AFRIKAANSE OPSOMMING: 18F-Fallypride is van waarde in neurologiese en psigiatriese siekte om dopamien D2/D3 reseptore in striatale en ekstrastriatale areas te visualiseer. Verskeie vereenvoudigde metodes vir kwantifisering van 18F-fallypride positron emissie tomografie (PET) breinbeelding, is deur middel van nie-ingrypende metodes op ware data beskryf en gebruik. Systemiese analise van die effek van hierdie metodes op kwantifisering is nog nie uitgevoer nie. In die huidige studie, is wiskundige simulasies gebruik om hierdie invloed op die kwantifisering te bestudeer, deur van verskillende metodes gebruik te maak om 18F-Fallypride PET in die menslike brein te kwantifiseer. Die spesifike areas van opname in hierdie studie, is die putamen (hoë reseptor digtheidsarea), talamus (matige reseptor digtheidsarea) en temporale korteks (lae digtheidsarea). Materiaal en metodes Alle simulasies is op Matlab (weergawe R2013a (8.1.0.604) win 64-bit sagteware; MathWorks, Inc.) uitgevoer. Die werklike metings is gesimuleer deur modellering met wisselende raamduur, verval van spoorelemente en wisselende geraasvlakke, gebasseer op ideale weefselkurwes. ‘n Gemodelleerde insetfunksie is gebruik vir die konstruksie van ideale weefselkurwes. Kwantifisering is uitgevoer met arteriële bloedmonster-modelle: ‘n 2T4k (twee-weefsel of drie-kompartement) omkeerbare model en ‘n Logan grafiese analise. Die serebellum is as verwysingsgebied gebruik, om die volgende modelle te bestudeer: ‘n Logan grafiese analise (Loganref), ‘n verwysingsweefsel-model (VWM) en die vereenvoudigde verwysingsweefsel-model (VVWM). Laastens is die standard-opname verhoudingswaarde as die eenvoudigste metode gebruik. Die resultate is geanaliseer deur van ‘n korrelasie-analise en ‘n Blandt-Altman analise gebruik te maak. Die relatiewe fout (%), sowel as die 5de en 95ste persentiel in hoë/middelmatige/lae reseptor digtheidsarea van die brein, is bepaal. Die geldigheid van elke metode is geëvalueer in terme van vooroordeel en variasie. Resultate Die studie het in alle gevalle getoon dat die grondwaarheidsmetode en die grafiese metode, wat van die arteriële insetfunksie gebruik maak, byna identies is. Die algehele akkuraatheid en die veranderlike van die verskillende metodes, was suksesvol. Loganref analise was die akkuraatste in die lae/middelmatige reseptor-digtheidsareas in plaas van die gebruik van die ingrypende arteriële bloedmonster metode. Die vereenvoudigde verwysingsweefsel-model (VVWM) was die akkuraatste in die hoë reseptor digtheidsarea, daarenteen het die standard-opname verhoudingswaarde relatief groot foute getoon. Die effek van sein tot geraasverhouding op kwatifisering was duidelik, aangesien die vooroordeel van alle verwysingsgebied-metodes met toenemende geraas verhoog het. Wanneer die geraasvlak egter te hoog is, word die vooroordeel ook te groot. ‘n Ondersoek na die effek van modelverrigting onder die aaneenlopende monsterprogram, toon stabiliteit in die kinetika van 18F-fallypride. Hierdie effek vind plaas met die gebruik van ‘n minder ingrypende verwysingsmetode wanneer die beeldingstyd > 150 minute en > 120 minute in onderskeidelik hoë en in lae/middelmatige reseptor- digtheidsareas is. Die beeldingtydsduur met tussensposes was gebruik om ‘n ooreenkoms tussen akkuraatheid en pasiëntgemak te bereik. Die beste kwantitatiewe resultate is verkry met die volgende program: 30 minute beelding, 30 minute tussenpose, gevolg deur 30 minute beelding met ‘n 120 minute onderbreking en laastens weer ‘n 30 minute beelding. Gevolgtrekking Vereenvoudigde modelle kan gebruik word om nuttige voorspellings van dopamien- draers, te verkry, wat vergelykbaar is met metodes waar arteriële bloedmonsters gebruik word. Die modelle moet egter met groot sorg gebruik word, aangesien dit deur datageraas, die duur van die beelding, die tipe onderbrekings of die onderbreking van die monsterprogram, beïnvloed kan word.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/105765
This item appears in the following collections: