NeuroImage 94 (2014) 408–410
Contents lists available at ScienceDirect
NeuroImage journal homepage: www.elsevier.com/locate/ynimg
Comments and Controversies
PET neuroimaging: The elephant unpacks his trunk Comment on Cumming: “PET neuroimaging: The white elephant packs his trunk?” Roger N. Gunn a,b,c,⁎, Eugenii A. Rabiner a,d a
Imanova, Centre for Imaging Sciences, London, UK Division of Brain Sciences, Department of Medicine, Imperial College, London, UK c Department of Engineering Science, University of Oxford, Oxford, UK d Centre for Neuroimaging Sciences, Institute of Psychiatry, King's College, London, UK b
In a recent NeuroImage commentary Paul Cumming uses biomedical literature data to make inferences about the scientific future of brain PET (Cumming, 2013). Using the search terms PET, fMRI, and Brain in PubMed, Cumming attempts to analyse the broader historical trends of brain PET imaging. Significant scepticism should be maintained at the use of crude measures, such as total publication number, to infer the value of a particular scientific field. Nevertheless, such methodology can provide a rough estimate of a scientific field's general activity (where to paraphrase Sir Humphrey Appleby1 “activity can serve as a substitute for achievement”). An initial examination of the data presented drew our attention to some obvious errors. For example the graph presented in the original commentary (Fig. 1) shows a large number of fMRI neuroimaging papers published in the early 1980s. To our knowledge the first human fMRI papers were published in 1992. In addition, the number of PET neuroimaging papers that the graph claims were published in 2012 is too low. Indeed, our group at Imanova, alone published a number of PET neuroimaging papers in 2012 similar to that indicated for the field as a whole. Whilst having a naturally high regard for our own work, we believe that others are successfully publishing in this field too. Nevertheless, we thought it would be interesting to take a closer look at the real data ourselves. We used Scopus (www.scopus.com)2 to search the Title/Abstract/Keywords category for the queries listed in Table 1. An examination of the output of this search (see Fig. 2) indicates that both PET and fMRI neuroimaging continue to grow in terms of the number of papers published per year. fMRI overtook PET, in terms of the total number of papers published per year, around 2002. In order to examine the historic trends and try to project them into the future (with the many significant caveats that extrapolation entails) we fitted these data to the following asymptotic equation, ln ðpapers=yearÞ ¼
αt tþβ
ð1Þ
⁎ Corresponding author at: Imanova, 2nd Floor, Burlington Danes Building, Imperial College London, Du Cane Road, London W12 0NN, UK. E-mail address: [email protected] (R.N. Gunn). 1 Sir Humphrey Appleby, GCB, KBE, MVO, MA (Oxon), is a fictional character from the British television series Yes Minister and Yes, Prime Minister. The quote “Politicians must be allowed to panic. They need activity. It is their substitute for achievement.” is taken from an episode called The Economy Drive, which perhaps aptly was recorded in1980. 2 Similar results were also obtained from PubMed. 1053-8119/$ – see front matter © 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.neuroimage.2013.12.042
where t is the time in years following the introduction of the technology (we chose 1979 for PET and 1992 for fMRI), while α and β are parameters to be estimated. In this model, α represents the asymptote of the ln(papers/year) and β is the number of years it takes each field to reach half its asymptotic value in log terms, and can thus be viewed as parameter which reflects the rate of adoption of the technology. The asymptote for fMRI (α = 9.60) is higher than for PET (α = 8.01) equating to an asymptotic publication rate of around 15,000 and 3000 papers/year for fMRI and PET respectively. Examination of the parameter β tells us that the adoption rate of fMRI was faster than PET (β = 4.66 years for fMRI and β = 5.81 years for PET), consistent with a simpler, cheaper and more accessible technology. PET neuroimaging research can be divided into a number of broad areas, such as investigations into normo-/pathophysiology, pharmacology (including drug development) and diagnostics. fMRI most closely relates to [15O]H2O PET methods used to study brain activation. We thus investigated the impact fMRI had on [15O]H2O PET (here we have taken the search for Oxygen-15 PET studies to be dominated by [15O]H2O PET with the understanding that there will be a small over-estimate due to the inclusion of other publications that pertain to blood volume and oxygen metabolism). The data presented in Fig. 3 demonstrate that Oxygen-15 PET studies peak around the year 2000 and decline subsequently with the maxima coinciding with the time at which the fMRI publication rate overtakes that of PET. The time courses suggest a pre-cursor product relationship, with PET blood flow studies establishing the principles of the field, encouraging the development of simpler methodologies, and then decaying as fMRI takes over (see Table 1 Search terms used to query the Scopus biomedical literature database for the number of published articles between 1980 and 2012. Search terms
Category of brain research
Brain AND DOCTYPE(ar) fMRI AND Brain AND DOCTYPE(ar) PET AND Brain AND DOCTYPE(ar) PET AND Brain AND (11C or C11 or 11-C or C-11) AND DOCTYPE(ar) PET AND Brain AND (18F or F18 or 18-F or F-18) AND DOCTYPE(ar) PET AND Brain AND (15O or O15 or 15-O or O-15) AND DOCTYPE(ar)
Brain studies fMRI studies PET studies Carbon-11 PET studies Fluorine-18 PET studies Oxygen-15 PET studies
R.N. Gunn, E.A. Rabiner / NeuroImage 94 (2014) 408–410
409
Fig. 3. PET neuroimaging. C11 and F18 PET neuroimaging studies continue to increase whilst O15 studies peaked around 2000 and are now declining. Fig. 1. Reproduction of figure illustrating the time course of fMRI and PET neuroimaging publications from Cumming (2013).
Fig. 4). In contrast, both PET neuroimaging as a whole, and C11 and F18 PET studies continue to increase year on year (Fig. 3), with an evergreater number of probes becoming available for CNS targets. The range of these targets is also increasing rapidly, with the early focus on G protein coupled receptors expanding to encompass transporters, enzymes and misfolded proteins. Thus, perhaps the real message is that the cheaper and simpler method of fMRI with its increased spatial resolution and ease of use has largely taken over from PET in the field of activation studies. However, areas of PET neuroimaging where cheaper and simpler MRI alternatives are not available continue to grow. The application of PET to CNS drug development has become almost de rigeur and it is a rare exception these days for a drug development programme to not include PET studies in the options considered. In the field of diagnostics, the imaging of neurodegenerative conditions using dopaminergic markers for Parkinson's disease, and β-amyloid and more recently tau-tracers for Alzheimer's disease, is growing rapidly. We believe that the two technologies have different strengths and that they typically complement each other rather than compete (see
Fig. 2. Number of papers published per year for Brain research, fMRI neuroimaging and PET neuroimaging. Neuroimaging data are fitted to the asymptotic function given in Eq. (1) (fits are shown as solid lines) which indicates a faster adoption rate for fMRI and larger publication rate from the year 2002 onwards.
for example Rabiner et al., 2011; Horwitz and Simonyan, 2013; Siebner et al., 2013). fMRI has higher spatial and temporal resolution, and has the flexibility to provide information on the CNS effects of a variety of compounds and stimuli. It often struggles however to dissect the specific contribution of a particular CNS target, or a particular molecule — an area where PET excels. PET also possesses unrivalled sensitivity with the ability to measure sub-picomolar concentrations. Finally, PET allows for an accurate quantification of the biological substrate underlying the signal, and an intuitive translation of the imaging signal into meaningful neurochemical and physiological parameters, which can require a lot more effort in fMRI. It should also be noted that MRI offers more to neuroscience than simply fMRI (by which people often mean just BOLD MRI). Computational anatomy, tractography and spectroscopy, to name but three, offer valuable avenues for the exploration of brain function and these can additionally support PET studies in multi-modal designs when appropriate (see for example Tziortzi et al., 2013). However, rather than exploring these areas here, we restricted ourselves to what we believe was the intended analyses of Cumming. Finally, any analysis of the biomedical literature should be viewed critically, with regard to questions such as: how important is the number
Fig. 4. Activation studies. PET as a precursor for fMRI — as fMRI is adopted aggressively PET activation studies peak and start to decline.
410
R.N. Gunn, E.A. Rabiner / NeuroImage 94 (2014) 408–410
of publications/year, how effective are the search terms and the search categories in identifying the right papers etc. Nevertheless, all things considered we believe the overall data presented here provide a useful overview to the trajectories of the technologies discussed.
References Cumming, P., 2014. PET neuroimaging: the white elephant packs his trunk? Neuroimaging 84, 1094–1100. Horwitz, B., Simonyan, K., 2014. PET neuroimaging: plenty of studies still need to be performed: comment on Cumming: “PET neuroimaging: the white elephant packs his trunk?”. Neuroimaging 84, 1101–1103.
Rabiner, E.A., Beaver, J., Makwana, A., Searle, G., Long, C., Nathan, P.J., Newbould, R.D., Howard, J., Miller, S.R., Bush, M.A., Hill, S., Reiley, R., Passchier, J., Gunn, R.N., Matthews, P.M., Bullmore, E.T., 2011. Pharmacological differentiation of opioid receptor antagonists by molecular and functional imaging of target occupancy and food reward-related brain activation in humans. Mol. Psychiatry 785 (8), 826–835. http:// dx.doi.org/10.1038/mp.2011.29. (Epub 2011 Apr 19. PubMed PMID: 21502953; PubMed Central PMCID: PMC3142667). Siebner, H.R., Strafella, A.P., Rowe, J.B., 2013. The white elephant revived: a new marriage between PET and MRI: comment to Cumming: “PET neuroimaging: the white elephant packs his trunk?”. Neuroimaging 84, 1104–1106. Tziortzi, A.C., Haber, S.N., Searle, G.E., Tsoumpas, C., Long, C.J., Shotbolt, P., Douaud, G., Jbabdi, S., Behrens, T.E., Rabiner, E.A., Jenkinson, M., Gunn, R.N., 2013. Connectivitybased functional analysis of dopamine release in the striatum using diffusionweighted MRI and positron emission tomography. (Jan 2.) Cereb. Cortex ([Epub ahead of print] PubMed PMID: 23283687).
Contents lists available at ScienceDirect
NeuroImage journal homepage: www.elsevier.com/locate/ynimg
Comments and Controversies
PET neuroimaging: The elephant unpacks his trunk Comment on Cumming: “PET neuroimaging: The white elephant packs his trunk?” Roger N. Gunn a,b,c,⁎, Eugenii A. Rabiner a,d a
Imanova, Centre for Imaging Sciences, London, UK Division of Brain Sciences, Department of Medicine, Imperial College, London, UK c Department of Engineering Science, University of Oxford, Oxford, UK d Centre for Neuroimaging Sciences, Institute of Psychiatry, King's College, London, UK b
In a recent NeuroImage commentary Paul Cumming uses biomedical literature data to make inferences about the scientific future of brain PET (Cumming, 2013). Using the search terms PET, fMRI, and Brain in PubMed, Cumming attempts to analyse the broader historical trends of brain PET imaging. Significant scepticism should be maintained at the use of crude measures, such as total publication number, to infer the value of a particular scientific field. Nevertheless, such methodology can provide a rough estimate of a scientific field's general activity (where to paraphrase Sir Humphrey Appleby1 “activity can serve as a substitute for achievement”). An initial examination of the data presented drew our attention to some obvious errors. For example the graph presented in the original commentary (Fig. 1) shows a large number of fMRI neuroimaging papers published in the early 1980s. To our knowledge the first human fMRI papers were published in 1992. In addition, the number of PET neuroimaging papers that the graph claims were published in 2012 is too low. Indeed, our group at Imanova, alone published a number of PET neuroimaging papers in 2012 similar to that indicated for the field as a whole. Whilst having a naturally high regard for our own work, we believe that others are successfully publishing in this field too. Nevertheless, we thought it would be interesting to take a closer look at the real data ourselves. We used Scopus (www.scopus.com)2 to search the Title/Abstract/Keywords category for the queries listed in Table 1. An examination of the output of this search (see Fig. 2) indicates that both PET and fMRI neuroimaging continue to grow in terms of the number of papers published per year. fMRI overtook PET, in terms of the total number of papers published per year, around 2002. In order to examine the historic trends and try to project them into the future (with the many significant caveats that extrapolation entails) we fitted these data to the following asymptotic equation, ln ðpapers=yearÞ ¼
αt tþβ
ð1Þ
⁎ Corresponding author at: Imanova, 2nd Floor, Burlington Danes Building, Imperial College London, Du Cane Road, London W12 0NN, UK. E-mail address: [email protected] (R.N. Gunn). 1 Sir Humphrey Appleby, GCB, KBE, MVO, MA (Oxon), is a fictional character from the British television series Yes Minister and Yes, Prime Minister. The quote “Politicians must be allowed to panic. They need activity. It is their substitute for achievement.” is taken from an episode called The Economy Drive, which perhaps aptly was recorded in1980. 2 Similar results were also obtained from PubMed. 1053-8119/$ – see front matter © 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.neuroimage.2013.12.042
where t is the time in years following the introduction of the technology (we chose 1979 for PET and 1992 for fMRI), while α and β are parameters to be estimated. In this model, α represents the asymptote of the ln(papers/year) and β is the number of years it takes each field to reach half its asymptotic value in log terms, and can thus be viewed as parameter which reflects the rate of adoption of the technology. The asymptote for fMRI (α = 9.60) is higher than for PET (α = 8.01) equating to an asymptotic publication rate of around 15,000 and 3000 papers/year for fMRI and PET respectively. Examination of the parameter β tells us that the adoption rate of fMRI was faster than PET (β = 4.66 years for fMRI and β = 5.81 years for PET), consistent with a simpler, cheaper and more accessible technology. PET neuroimaging research can be divided into a number of broad areas, such as investigations into normo-/pathophysiology, pharmacology (including drug development) and diagnostics. fMRI most closely relates to [15O]H2O PET methods used to study brain activation. We thus investigated the impact fMRI had on [15O]H2O PET (here we have taken the search for Oxygen-15 PET studies to be dominated by [15O]H2O PET with the understanding that there will be a small over-estimate due to the inclusion of other publications that pertain to blood volume and oxygen metabolism). The data presented in Fig. 3 demonstrate that Oxygen-15 PET studies peak around the year 2000 and decline subsequently with the maxima coinciding with the time at which the fMRI publication rate overtakes that of PET. The time courses suggest a pre-cursor product relationship, with PET blood flow studies establishing the principles of the field, encouraging the development of simpler methodologies, and then decaying as fMRI takes over (see Table 1 Search terms used to query the Scopus biomedical literature database for the number of published articles between 1980 and 2012. Search terms
Category of brain research
Brain AND DOCTYPE(ar) fMRI AND Brain AND DOCTYPE(ar) PET AND Brain AND DOCTYPE(ar) PET AND Brain AND (11C or C11 or 11-C or C-11) AND DOCTYPE(ar) PET AND Brain AND (18F or F18 or 18-F or F-18) AND DOCTYPE(ar) PET AND Brain AND (15O or O15 or 15-O or O-15) AND DOCTYPE(ar)
Brain studies fMRI studies PET studies Carbon-11 PET studies Fluorine-18 PET studies Oxygen-15 PET studies
R.N. Gunn, E.A. Rabiner / NeuroImage 94 (2014) 408–410
409
Fig. 3. PET neuroimaging. C11 and F18 PET neuroimaging studies continue to increase whilst O15 studies peaked around 2000 and are now declining. Fig. 1. Reproduction of figure illustrating the time course of fMRI and PET neuroimaging publications from Cumming (2013).
Fig. 4). In contrast, both PET neuroimaging as a whole, and C11 and F18 PET studies continue to increase year on year (Fig. 3), with an evergreater number of probes becoming available for CNS targets. The range of these targets is also increasing rapidly, with the early focus on G protein coupled receptors expanding to encompass transporters, enzymes and misfolded proteins. Thus, perhaps the real message is that the cheaper and simpler method of fMRI with its increased spatial resolution and ease of use has largely taken over from PET in the field of activation studies. However, areas of PET neuroimaging where cheaper and simpler MRI alternatives are not available continue to grow. The application of PET to CNS drug development has become almost de rigeur and it is a rare exception these days for a drug development programme to not include PET studies in the options considered. In the field of diagnostics, the imaging of neurodegenerative conditions using dopaminergic markers for Parkinson's disease, and β-amyloid and more recently tau-tracers for Alzheimer's disease, is growing rapidly. We believe that the two technologies have different strengths and that they typically complement each other rather than compete (see
Fig. 2. Number of papers published per year for Brain research, fMRI neuroimaging and PET neuroimaging. Neuroimaging data are fitted to the asymptotic function given in Eq. (1) (fits are shown as solid lines) which indicates a faster adoption rate for fMRI and larger publication rate from the year 2002 onwards.
for example Rabiner et al., 2011; Horwitz and Simonyan, 2013; Siebner et al., 2013). fMRI has higher spatial and temporal resolution, and has the flexibility to provide information on the CNS effects of a variety of compounds and stimuli. It often struggles however to dissect the specific contribution of a particular CNS target, or a particular molecule — an area where PET excels. PET also possesses unrivalled sensitivity with the ability to measure sub-picomolar concentrations. Finally, PET allows for an accurate quantification of the biological substrate underlying the signal, and an intuitive translation of the imaging signal into meaningful neurochemical and physiological parameters, which can require a lot more effort in fMRI. It should also be noted that MRI offers more to neuroscience than simply fMRI (by which people often mean just BOLD MRI). Computational anatomy, tractography and spectroscopy, to name but three, offer valuable avenues for the exploration of brain function and these can additionally support PET studies in multi-modal designs when appropriate (see for example Tziortzi et al., 2013). However, rather than exploring these areas here, we restricted ourselves to what we believe was the intended analyses of Cumming. Finally, any analysis of the biomedical literature should be viewed critically, with regard to questions such as: how important is the number
Fig. 4. Activation studies. PET as a precursor for fMRI — as fMRI is adopted aggressively PET activation studies peak and start to decline.
410
R.N. Gunn, E.A. Rabiner / NeuroImage 94 (2014) 408–410
of publications/year, how effective are the search terms and the search categories in identifying the right papers etc. Nevertheless, all things considered we believe the overall data presented here provide a useful overview to the trajectories of the technologies discussed.
References Cumming, P., 2014. PET neuroimaging: the white elephant packs his trunk? Neuroimaging 84, 1094–1100. Horwitz, B., Simonyan, K., 2014. PET neuroimaging: plenty of studies still need to be performed: comment on Cumming: “PET neuroimaging: the white elephant packs his trunk?”. Neuroimaging 84, 1101–1103.
Rabiner, E.A., Beaver, J., Makwana, A., Searle, G., Long, C., Nathan, P.J., Newbould, R.D., Howard, J., Miller, S.R., Bush, M.A., Hill, S., Reiley, R., Passchier, J., Gunn, R.N., Matthews, P.M., Bullmore, E.T., 2011. Pharmacological differentiation of opioid receptor antagonists by molecular and functional imaging of target occupancy and food reward-related brain activation in humans. Mol. Psychiatry 785 (8), 826–835. http:// dx.doi.org/10.1038/mp.2011.29. (Epub 2011 Apr 19. PubMed PMID: 21502953; PubMed Central PMCID: PMC3142667). Siebner, H.R., Strafella, A.P., Rowe, J.B., 2013. The white elephant revived: a new marriage between PET and MRI: comment to Cumming: “PET neuroimaging: the white elephant packs his trunk?”. Neuroimaging 84, 1104–1106. Tziortzi, A.C., Haber, S.N., Searle, G.E., Tsoumpas, C., Long, C.J., Shotbolt, P., Douaud, G., Jbabdi, S., Behrens, T.E., Rabiner, E.A., Jenkinson, M., Gunn, R.N., 2013. Connectivitybased functional analysis of dopamine release in the striatum using diffusionweighted MRI and positron emission tomography. (Jan 2.) Cereb. Cortex ([Epub ahead of print] PubMed PMID: 23283687).
