Abstract Along with the prevalence of movement disorders, the use of 123 I-labeled 2b-carbomethoxy-3b-(4-iodophenyl)-N-(3-fluoropropyl)nortropane ( 123 I- FP-CIT) single photon emission computed tomography (SPECT) as a diagnosis tool has increased. Several softwares have been created in order to improve quantification of relative nigrostriatal uptake so that it can be used as a reliable adjunct to visual assessment. Our aim was to prospectively evaluate the influence of automatic, semiautomatic and manual algorithms on DaTSCAN uptake semiquantification. DaTSCAN quantification: comparative analysis of different processing algorithms Rui Miguel Correia 1 ; Luísa Roldão Pereira 1 ; Diogo Borges Faria 2,3 ; Lina Vieira 4,5 ; Joana Vale 2,3 ; Andreia Martins 2 ; Joana Patrina 2 ; Joana Fernandes 2 ; Joana Teixeira 2 ; Diogo Sousa 2 ; José Manuel Oliveira 2 1 Nuclear Medicine degree, Escola Superior de Tecnologia da Saúde de Lisboa, Instituto Politécnico de Lisboa, Portugal; 2 HPP – Medicina Molecular, SA; Lenitudes SGPS, SA; 3 ESSUA – Escola Superior de Saúde da Universidade de Aveiro; 4 Área Científica de Medicina Nuclear, Escola Superior de Tecnologia da Saúde de Lisboa, Instituto Politécnico de Lisboa, Portugal; 5 Instituto de Biofísica e Engenharia Biomédica, Faculdade de Ciências, Universidade de Lisboa, Lisboa, Portugal Introduction Over the last decade, imaging techniques for probing brain function had a great development, especially since neuroprotective strategies have become available. Striatal single photon emission computed tomography (SPECT) allows the perception of the nigrostriatal pathway where degeneration is known to be a source of movement disorders, such as multiple system atrophy (MSA) (1) and Parkinson’s disease (PD) (2) . Image evaluation is traditionally done by visual analysis with no significant prejudice in terms of sensibility and specificity in the identification of loss of striatal presynaptic dopamine transporters (DAT) (4,5) . Semiquantification has proved itself as a beneficial and promising tool, since it offers a less subjective adjunct to visual interpretation (3) , by calculation of ratios that express the relation between an area of specific uptake ( 123 I-FP-CIT uptake in the striatum) and a region of nonspecific uptake – a background region devoid of DATs (2,7,8) . The algorithms can be classified as fully manual, semiautomatic and automatic. Costa’s Method (CM), QuantiSPECT-3 Boxes (QS3B) and Computed Solution for Automated Quantification (CSAQ) have the same principle, differing only in the level of user interaction. Southampton’s Method (SM) and QuantiSPECT-2 Boxes (QS2B) follows the same direction. DatQUANT is the exception, with an anatomical VOI and 2 VOI’s on occipital as background. A B C D E F Figure 1. Final display of each algorithm: (A) CM, (B) QS3B, (C) CSAQ, (D) SM, (E) QS2B and (F) DaTQUANT. Conclusions Exception made for CSAQ’s negative predictive value and CM’s specificity, these, along with DaTQUANT seem to be the more consistent algorithms with the visual evaluation. Automatic algorithms seem a trustworthy time-saving time option. Quantification method choice is clinically relevant in the delivery of reliable patient care. .Spect I, Djang DSW, Janssen MJR, Bohnen N, Booij J, Henderson TA, et al. SNM Practice Guideline for Dopamine Transporter Imaging. 2012;53(1):154–63. .Lucena, Filipa; Sousa, Eva; F. Vaz T. Imaging in neurological and vascular brain diseases (SPECT and SPECT/CT). In: Fragoso Costa, Pedro; Santos, Andrea;Vidovič B, editor. Brain Imaging - Technologist’s Guide. 1st ed. EANM; 2015. p. 54–71. .Tossici-bolt L, Hoffmann SMA, Kemp PM, Mehta RL, Fleming JS. Original article Quantification of [ 123 I ] FP-CIT SPECT brain images : an accurate technique for measurement of the specific binding ratio. 2006;33(12). .Augimeri A, Cherubini A, Cascini GL, Galea D, Caligiuri ME, Barbagallo G, et al. CADA — computer-aided DaTSCAN analysis. EJNMMI Phys [Internet]. EJNMMI Physics; 2016; Available from: http://dx.doi.org/10.1186/s40658-016-0140-9 .Pagan, Laura; Tranfaglia, C; Galli, S; Lucchi, G; Novi, B; Fagioli, G; Guidarelli G. A comparison between the conventional manual ROI method and an automatic algorithm for semiquantitative analysis of SPECT studies. In: Journal ofPhysics:ConferenceSeries. IOP Publishing; 2011. .Declerck, Ralph Buchert; Chloe Hutton; Lange, Catharina ; Hoppe, Peter; Makowski, Marcus; Bamousa, Thamer; Platsch, Gunther; Brenner W. Semiquantitative slab view display for visual evaluation of 123I-FP-CIT SPECT. Nucl Med Commun. 2016;37:509–18. .Wang L, Zhang Q, Li H, Zhang H. SPECT Molecular Imaging in Parkinson ’ s Disease. J Biomed Biotechnol. 2012;2012. References Methods Results Table 2. Values obtained from ROC analysis. The best values of each category are presented with gold background. Friedman test presented p values of 0.13 (significant threshold was set at p=0.05). Discussion Qualitative analysis had no significant difference from visual analysis (p=0.13). Friedman test revealed that there are statistically significant differences, specifically between the manual and the automatic methods. Southampton method persistently presents the less corresponding results. ROC curve played a crucial role in evaluating diagnostic ability of algorithms. In general, semiquantitative algorithms provide equivalent diagnostic accuracy as visual analysis. With no other correction besides Chang’s, SUB differences may have been amplified, subsequently affecting the categorical classification. SPECT images suffer from poor spatial resolution, which leads to partial volume effect (PVE) due to cross-talk between different anatomical regions. When no correction is applied to this parameter, the measured SUB concentration is reduced occasioning lower SUB values. DaTQUANT ponders the age effect, while the others may overestimate the level of degeneration by presenting a value exacerbated by the normal SUB age-decline. Table 3. Values from Pearson Correlation and Youden Index for each algorithm. Sample 37 patients who underwent DaTSCAN due to movement disorders specifically PD suspicion. Acquisition Protocol SPECT studies acquired according to EANM guidelines, on dual-head cameras system (Infinia Hawkeye or Millennium MG - GE Healthcare). Processing The imaging data was processed by the 6 algorithms. VOIs were applied to 3D maps to extract mean VOI scores, and to generate regional numeric indices for specific uptake binding (SUB). Statistical Analysis According to a known threshold, the SUB values were categorized as normal (=0)/ abnormal (=1) for each method and patient and used to make comparisons between each method and in relation to two experts’ visual analysis (figure 1). Pearson correlation, receiver operating characteristic (ROC) curve analysis, Friedman test and Wilcoxon test were calculated. [email protected]
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Abstract Along with the prevalence of movement disorders, the use of 123I-labeled 2b-carbomethoxy-3b-(4-iodophenyl)-N-(3-fluoropropyl)nortropane (123I-FP-CIT) single photon emission computed tomography (SPECT) as a diagnosis tool has increased. Several softwares have been created in orderto improve quantification of relative nigrostriatal uptake so that it can be used as a reliable adjunct to visual assessment.
Our aim was to prospectively evaluate the influence of automatic, semiautomatic and manual algorithms on DaTSCAN uptake semiquantification.
DaTSCAN quantification: comparative analysis of different processing algorithmsRui Miguel Correia1; Luísa Roldão Pereira1; Diogo Borges Faria2,3; Lina Vieira4,5; Joana Vale2,3; Andreia Martins2;
Joana Patrina2; Joana Fernandes2; Joana Teixeira2; Diogo Sousa2; José Manuel Oliveira2
1 Nuclear Medicine degree, Escola Superior de Tecnologia da Saúde de Lisboa, Instituto Politécnico de Lisboa, Portugal;2 HPP – Medicina Molecular, SA; Lenitudes SGPS, SA;3 ESSUA – Escola Superior de Saúde da Universidade de Aveiro;4 Área Científica de Medicina Nuclear, Escola Superior de Tecnologia da Saúde de Lisboa, Instituto Politécnico de Lisboa, Portugal;
5 Instituto de Biofísica e Engenharia Biomédica, Faculdade de Ciências, Universidade de Lisboa, Lisboa, Portugal
Introduction
Over the last decade, imaging techniques for probing brain function had a greatdevelopment, especially since neuroprotective strategies have become available.Striatal single photon emission computed tomography (SPECT) allows the perception ofthe nigrostriatal pathway where degeneration is known to be a source of movementdisorders, such as multiple system atrophy (MSA)(1) and Parkinson’s disease (PD)(2).Image evaluation is traditionally done by visual analysis with no significant prejudice interms of sensibility and specificity in the identification of loss of striatal presynapticdopamine transporters (DAT)(4,5).Semiquantification has proved itself as a beneficial and promising tool, since it offers aless subjective adjunct to visual interpretation(3), by calculation of ratios that express therelation between an area of specific uptake (123I-FP-CIT uptake in the striatum) and aregion of nonspecific uptake – a background region devoid of DATs(2,7,8).The algorithms can be classified as fully manual, semiautomatic and automatic.
Costa’s Method (CM), QuantiSPECT-3 Boxes (QS3B) and Computed Solution forAutomated Quantification (CSAQ) have the same principle, differing only in the level ofuser interaction. Southampton’s Method (SM) and QuantiSPECT-2 Boxes (QS2B)follows the same direction. DatQUANT is the exception, with an anatomical VOI and 2VOI’s on occipital as background.
A B C
D E F
Figure 1. Final display of each algorithm: (A) CM, (B) QS3B, (C) CSAQ, (D) SM, (E) QS2B and (F) DaTQUANT.
Conclusions
Exception made for CSAQ’s negative predictive value and CM’s specificity, these, along with DaTQUANT seem to be the more consistent algorithms with the visual evaluation. Automatic algorithms seem a trustworthy time-saving time option.Quantification method choice is clinically relevant in the delivery of reliable patient care.
.Spect I, Djang DSW, Janssen MJR, Bohnen N, Booij J, Henderson TA, et al. SNM Practice Guideline for Dopamine Transporter Imaging. 2012;53(1):154–63.
.Lucena, Filipa; Sousa, Eva; F. Vaz T. Imaging in neurological and vascular brain diseases (SPECT and SPECT/CT). In: Fragoso Costa, Pedro; Santos, Andrea;Vidovič B, editor. Brain Imaging - Technologist’s Guide. 1st ed. EANM; 2015. p. 54–71.
.Tossici-bolt L, Hoffmann SMA, Kemp PM, Mehta RL, Fleming JS. Original article Quantification of [ 123 I ] FP-CIT SPECT brain images : an accurate technique for measurement of the specific binding ratio. 2006;33(12).
.Augimeri A, Cherubini A, Cascini GL, Galea D, Caligiuri ME, Barbagallo G, et al. CADA — computer-aided DaTSCAN analysis. EJNMMI Phys [Internet]. EJNMMI Physics; 2016; Available from: http://dx.doi.org/10.1186/s40658-016-0140-9
.Pagan, Laura; Tranfaglia, C; Galli, S; Lucchi, G; Novi, B; Fagioli, G; Guidarelli G. A comparison between the conventional manual ROI method and an automatic algorithm for semiquantitative analysis of SPECT studies. In: Journal ofPhysics:ConferenceSeries. IOP Publishing; 2011.
.Declerck, Ralph Buchert; Chloe Hutton; Lange, Catharina ; Hoppe, Peter; Makowski, Marcus; Bamousa, Thamer; Platsch, Gunther; Brenner W. Semiquantitative slab view display for visual evaluation of 123I-FP-CIT SPECT. Nucl Med Commun. 2016;37:509–18.
.Wang L, Zhang Q, Li H, Zhang H. SPECT Molecular Imaging in Parkinson ’ s Disease. J Biomed Biotechnol. 2012;2012.
References
Methods
ResultsTable 2. Values obtained from ROC analysis. The best values of each category are presented with gold background.
Friedman test presented p values of 0.13 (significant threshold was set at p=0.05).
Discussion
Qualitative analysis had no significant difference from visual analysis (p=0.13).Friedman test revealed that there are statistically significant differences, specificallybetween the manual and the automatic methods.Southampton method persistently presents the less corresponding results.ROC curve played a crucial role in evaluating diagnostic ability of algorithms.In general, semiquantitative algorithms provide equivalent diagnostic accuracy asvisual analysis.With no other correction besides Chang’s, SUB differences may have beenamplified, subsequently affecting the categorical classification. SPECT imagessuffer from poor spatial resolution, which leads to partial volume effect (PVE) due tocross-talk between different anatomical regions. When no correction is applied tothis parameter, the measured SUB concentration is reduced occasioning lowerSUB values.DaTQUANT ponders the age effect, while the others may overestimate the level ofdegeneration by presenting a value exacerbated by the normal SUB age-decline.
Table 3. Values from Pearson Correlation and Youden Index for each algorithm.
Sample 37 patients who underwent DaTSCAN due to movement disordersspecifically PD suspicion.
AcquisitionProtocol
SPECT studies acquired according to EANM guidelines, on dual-headcameras system (Infinia Hawkeye or Millennium MG - GE Healthcare).
Processing The imaging data was processed by the 6 algorithms. VOIs were appliedto 3D maps to extract mean VOI scores, and to generate regional numericindices for specific uptake binding (SUB).
StatisticalAnalysis
According to a known threshold, the SUB values were categorized asnormal (=0)/ abnormal (=1) for each method and patient and used to makecomparisons between each method and in relation to two experts’ visualanalysis (figure 1).Pearson correlation, receiver operating characteristic (ROC) curveanalysis, Friedman test and Wilcoxon test were calculated.
Related Documents
