- Email: [email protected]
IC-P2-133: Reproducibility of automated ROI analysis of PET amyloid ligand [11C]PIB uptake
T60 IC-P2-133
Alzheimer’s Imaging Consortium IC-P2: Poster Presentations REPRODUCIBILITY OF AUTOMATED ROI ANALYSIS OF PET AMYLOID LIGAND [11C]PIB UPTAKE
Noora M. Scheinin1, Sargo Aalto2, Ian Wilson3, Nina Kemppainen1, Kjell Någren1, Marita Kailaja¨rvi4, Mika Scheinin4, Juha O. Rinne1, 1 Turku PET Centre, Turku, Finland; 2Turku PET Centre; Department of Psychology, Åbo Akademi University, Turku, Finland; 3Turku Imanet Oy, GE Healthcare, Turku, Finland; 4Department of Pharmacology and Clinical Research Services Turku (CRST), University of Turku, Turku, Finland. Contact e-mail: [email protected] Background: Positron emission tomography (PET) with [11C]PIB has been used to quantitate and visualize the accumulation of amyloid beta in Alzheimer’s disease (AD). Traditional manual region of interest (ROI) analysis is laborious and liable to operator-induced error. To avoid these drawbacks, we employed automated ROI analysis. Objective: The aim of this study was to evaluate the reliability of an automated image analysis technique including fitting of imaging data on the standard brain space and automated ROI analysis in a test-retest setting with repeated [11C]PIB scans. Methods: Six AD patients (4 women and 2 men, mean age 71.2 years, mean MMSE score 24.7) and four healthy controls (3 women and 1 man, mean age 68.5 years, mean MMSE score 28.2) were scanned twice with an average interval of six weeks. In order to enable short imaging sessions, convenient for patients, the accumulation of [11C]PIB was quantitated as region-to-cerebellum ratios during only 30 min of data acquisition (from 60 to 90 min after tracer injection). Spatial normalization of individual images onto the standard brain space was performed using Statistical Parametric Mapping (SPM) and a ligand-specific [11C]PIB template. Standardized ROIs were defined on MRI template images representing brain anatomy in accordance with the MNI (Montreal Neurological Institute) space. To assess the reliability of the [11C]PIB uptake measurement, absolute variability (VAR) values and intra-class correlation coefficients (ICC) were calculated. Results: The VAR values varied from 3.3 % to 5.3 % in AD subjects and from 2.7 % to 6.4 % in controls, indicating good to excellent reproducibility. The ICC was 0.78 - 0.95 in AD subjects and 0.47 - 0.84 in controls. Conclusions: Automated image analysis based on a short and cost-effective scanning protocol was reliable and sufficiently reproducible for studies aiming at quantitation and visualization of amyloid accumulation during progression of AD and of treatment effects. IC-P2-134
FDG-PET STUDIES IN EARLY-ONSET FAMILIAL ALZHEIMER’S DISEASE
Michael Scho¨ll1, Elka Stefanova2, Ove Almkvist3, Karin Axelman4, Anders Wall5, Bengt Långstro¨m5, Lars Lannfelt6, Matti Viitanen7, Agneta Nordberg1, 1Division of Alzheimer Neurobiology, Department of Neurobiology, Care Sciences and Society, Karolinska Institutet, Stockholm, Sweden; 2Institute of Neurology, Clinical Center of Serbia, Belgrade, Serbia; 3Department of Neurobiology, Care Sciences and Society, Karolinska Institutet, Stockholm, Sweden; 4Department of Geriatric Medicine, Karolinska University Hospital Huddinge, Stockholm, Sweden; 5Uppsala Imanet AB, Imanet, GE Healthcare, Uppsala, Sweden; 6Department of Public Health/Geriatrics, Uppsala University Hospital, Uppsala, Sweden; 7Division of Clinical Geriatrics, Department of Neurobiology, Care Sciences and Society, Karolinska Institutet, Stockholm, Sweden. Contact e-mail: [email protected] Background: Carriers of mutations that lead to the rare early-onset familial form of Alzheimer’s disease (eoFAD) represent a population in which pathological changes eventually causing the disease can be studied most early. This study aims to determine very early abnormalities and specific patterns in cerebral glucose metabolism (CMRglc) in families harbouring mutations in the PS1 and in the APP gene, as well as the relationship with cognitive impairment. Methods: Six related presymptomatic His163Tyr PS1 mutation carriers, four related presymptomatic carriers of an APPswe 670/671 gene mutation and a control group of 23 healthy non-carriers
underwent FDG-PET scanning. The His163Tyr mutation carriers were followed-up after two years. Another family comprising two Met146Val PS1 mutation carriers, of which one was demented when enrolled and the other developed AD during the study, and two non-mutation carriers was studied by FDG-PET over a period of four years. FDG-PET data were analysed either by Statistical Parametric Mapping (SPM) and/or Region of Interest (ROI) analysis. Neuropsychological tests were performed in conjunction with all PET examinations. Results: The six presymptomatic His163Tyr PS1 mutation carriers showed at baseline significantly decreased CMRglc in right thalamus as measured by ROI analysis in comparison to the non-carriers. SPM analysis revealed a tendency towards decreased CMRglc in this region. At follow-up this had evolved into a significantly declined CMRglc in right thalamus, shown by both ROI and SPM analysis. No impairment in cognitive performance was observed at baseline or after two years. The four presymptomatic APPswe 670/671 mutation carriers showed in a preliminary comparison with the non-carriers a strong tendency to lower CMRglc in right posterior cingulated, right frontal, and right and left parietotemporal cortex. In the family with the Met146Val PS1 mutation the demented carrier showed markedly reduced CMRglc in nearly all cerebral brain regions while the second carrier showed progressive decline in CMRglc mainly in parietal brain regions, commencing before the onset of cognitive dysfunction. Conclusions: FDG-PET enables detection of functional abnormalities before manifestation of clinical symptoms and cognitive deficits while different eoFAD mutations might account for diverse pathologic metabolic patterns at different stages of AD development. IC-P2-135
TENSOR BASED MORPHOMETRY TEST RETEST VALIDATION IN AD SUBJECTS
Matthew L. Senjem, Clifford Jack, Brad Boeve, David Knopman, Ron Petersen, Mayo Clinic, Rochester, MN, USA. Contact e-mail: [email protected] Background: Tensor -based morphometry (TBM) provides a method to probe inter-group differences in rates of decline of gray matter (GM) volume captured by serial MRI studies. Our objectives were to implement an automated TBM pipeline in SPM5, and validate the implementation by assessing test - retest reproducibility using independent samples of AD subjects. Methods: We identified 94 probable AD subjects from the Mayo Alzheimer’s Disease Research Center and Alzheimer’s Disease Patient Registry. Subjects had undergone serial T1 weighted 3-D volumetric 1.5T MRI examinations and were matched for age, gender, and inter-scan interval. The basic preprocessing steps of TBM are: 1) Compute deformation from late to early image. 2) Create softmean of early and deformedlate images. 3) Create image of Jacobian determinants. 4) Normalize and segment the soft-mean, applying the normalization to Jacobian image. 5) Multiply normalized Jacobian with GM. 6) Modulate and smooth the normalized early and late GM images. 8) Perform statistics. We split the 94 AD subjects into test and retest groups, with 47 subjects in each group. We computed voxel-wise change in GM density over serial scans in the AD test group and repeated this in the AD retest group. We then compared results for consistency in the test and retest AD groups. Results: Results are in agreement with the well established spatial distribution of neurofibrillary AD pathology. Rates of GM density loss were signifiacnt in the medial and basal temporal lobe, posterior cingulate/precuneus, lateral temporal-parietal neocortex, and subfrontal regions. Figures 1 and 2 are maps of GM loss in the AD test and AD retest groups, respectively (p⬍0.05, FDR corrected). Little difference is seen in the topography of rate of GM loss between the completely independent test and retest groups of AD subjects. A test vs. retest difference map showed no significant voxels, thus illustrating the stability of the TBM method. Conclusions: We implemented an automated TBM algorithm in SPM5. The algorithm demonstrated excellent test-retest reproducibility in independent subject samples. The maps displaying rates of atrophy agree quite well with the expected anatomic distribution of greatest neuronal pathology in AD.
Alzheimer’s Imaging Consortium IC-P2: Poster Presentations REPRODUCIBILITY OF AUTOMATED ROI ANALYSIS OF PET AMYLOID LIGAND [11C]PIB UPTAKE
Noora M. Scheinin1, Sargo Aalto2, Ian Wilson3, Nina Kemppainen1, Kjell Någren1, Marita Kailaja¨rvi4, Mika Scheinin4, Juha O. Rinne1, 1 Turku PET Centre, Turku, Finland; 2Turku PET Centre; Department of Psychology, Åbo Akademi University, Turku, Finland; 3Turku Imanet Oy, GE Healthcare, Turku, Finland; 4Department of Pharmacology and Clinical Research Services Turku (CRST), University of Turku, Turku, Finland. Contact e-mail: [email protected] Background: Positron emission tomography (PET) with [11C]PIB has been used to quantitate and visualize the accumulation of amyloid beta in Alzheimer’s disease (AD). Traditional manual region of interest (ROI) analysis is laborious and liable to operator-induced error. To avoid these drawbacks, we employed automated ROI analysis. Objective: The aim of this study was to evaluate the reliability of an automated image analysis technique including fitting of imaging data on the standard brain space and automated ROI analysis in a test-retest setting with repeated [11C]PIB scans. Methods: Six AD patients (4 women and 2 men, mean age 71.2 years, mean MMSE score 24.7) and four healthy controls (3 women and 1 man, mean age 68.5 years, mean MMSE score 28.2) were scanned twice with an average interval of six weeks. In order to enable short imaging sessions, convenient for patients, the accumulation of [11C]PIB was quantitated as region-to-cerebellum ratios during only 30 min of data acquisition (from 60 to 90 min after tracer injection). Spatial normalization of individual images onto the standard brain space was performed using Statistical Parametric Mapping (SPM) and a ligand-specific [11C]PIB template. Standardized ROIs were defined on MRI template images representing brain anatomy in accordance with the MNI (Montreal Neurological Institute) space. To assess the reliability of the [11C]PIB uptake measurement, absolute variability (VAR) values and intra-class correlation coefficients (ICC) were calculated. Results: The VAR values varied from 3.3 % to 5.3 % in AD subjects and from 2.7 % to 6.4 % in controls, indicating good to excellent reproducibility. The ICC was 0.78 - 0.95 in AD subjects and 0.47 - 0.84 in controls. Conclusions: Automated image analysis based on a short and cost-effective scanning protocol was reliable and sufficiently reproducible for studies aiming at quantitation and visualization of amyloid accumulation during progression of AD and of treatment effects. IC-P2-134
FDG-PET STUDIES IN EARLY-ONSET FAMILIAL ALZHEIMER’S DISEASE
Michael Scho¨ll1, Elka Stefanova2, Ove Almkvist3, Karin Axelman4, Anders Wall5, Bengt Långstro¨m5, Lars Lannfelt6, Matti Viitanen7, Agneta Nordberg1, 1Division of Alzheimer Neurobiology, Department of Neurobiology, Care Sciences and Society, Karolinska Institutet, Stockholm, Sweden; 2Institute of Neurology, Clinical Center of Serbia, Belgrade, Serbia; 3Department of Neurobiology, Care Sciences and Society, Karolinska Institutet, Stockholm, Sweden; 4Department of Geriatric Medicine, Karolinska University Hospital Huddinge, Stockholm, Sweden; 5Uppsala Imanet AB, Imanet, GE Healthcare, Uppsala, Sweden; 6Department of Public Health/Geriatrics, Uppsala University Hospital, Uppsala, Sweden; 7Division of Clinical Geriatrics, Department of Neurobiology, Care Sciences and Society, Karolinska Institutet, Stockholm, Sweden. Contact e-mail: [email protected] Background: Carriers of mutations that lead to the rare early-onset familial form of Alzheimer’s disease (eoFAD) represent a population in which pathological changes eventually causing the disease can be studied most early. This study aims to determine very early abnormalities and specific patterns in cerebral glucose metabolism (CMRglc) in families harbouring mutations in the PS1 and in the APP gene, as well as the relationship with cognitive impairment. Methods: Six related presymptomatic His163Tyr PS1 mutation carriers, four related presymptomatic carriers of an APPswe 670/671 gene mutation and a control group of 23 healthy non-carriers
underwent FDG-PET scanning. The His163Tyr mutation carriers were followed-up after two years. Another family comprising two Met146Val PS1 mutation carriers, of which one was demented when enrolled and the other developed AD during the study, and two non-mutation carriers was studied by FDG-PET over a period of four years. FDG-PET data were analysed either by Statistical Parametric Mapping (SPM) and/or Region of Interest (ROI) analysis. Neuropsychological tests were performed in conjunction with all PET examinations. Results: The six presymptomatic His163Tyr PS1 mutation carriers showed at baseline significantly decreased CMRglc in right thalamus as measured by ROI analysis in comparison to the non-carriers. SPM analysis revealed a tendency towards decreased CMRglc in this region. At follow-up this had evolved into a significantly declined CMRglc in right thalamus, shown by both ROI and SPM analysis. No impairment in cognitive performance was observed at baseline or after two years. The four presymptomatic APPswe 670/671 mutation carriers showed in a preliminary comparison with the non-carriers a strong tendency to lower CMRglc in right posterior cingulated, right frontal, and right and left parietotemporal cortex. In the family with the Met146Val PS1 mutation the demented carrier showed markedly reduced CMRglc in nearly all cerebral brain regions while the second carrier showed progressive decline in CMRglc mainly in parietal brain regions, commencing before the onset of cognitive dysfunction. Conclusions: FDG-PET enables detection of functional abnormalities before manifestation of clinical symptoms and cognitive deficits while different eoFAD mutations might account for diverse pathologic metabolic patterns at different stages of AD development. IC-P2-135
TENSOR BASED MORPHOMETRY TEST RETEST VALIDATION IN AD SUBJECTS
Matthew L. Senjem, Clifford Jack, Brad Boeve, David Knopman, Ron Petersen, Mayo Clinic, Rochester, MN, USA. Contact e-mail: [email protected] Background: Tensor -based morphometry (TBM) provides a method to probe inter-group differences in rates of decline of gray matter (GM) volume captured by serial MRI studies. Our objectives were to implement an automated TBM pipeline in SPM5, and validate the implementation by assessing test - retest reproducibility using independent samples of AD subjects. Methods: We identified 94 probable AD subjects from the Mayo Alzheimer’s Disease Research Center and Alzheimer’s Disease Patient Registry. Subjects had undergone serial T1 weighted 3-D volumetric 1.5T MRI examinations and were matched for age, gender, and inter-scan interval. The basic preprocessing steps of TBM are: 1) Compute deformation from late to early image. 2) Create softmean of early and deformedlate images. 3) Create image of Jacobian determinants. 4) Normalize and segment the soft-mean, applying the normalization to Jacobian image. 5) Multiply normalized Jacobian with GM. 6) Modulate and smooth the normalized early and late GM images. 8) Perform statistics. We split the 94 AD subjects into test and retest groups, with 47 subjects in each group. We computed voxel-wise change in GM density over serial scans in the AD test group and repeated this in the AD retest group. We then compared results for consistency in the test and retest AD groups. Results: Results are in agreement with the well established spatial distribution of neurofibrillary AD pathology. Rates of GM density loss were signifiacnt in the medial and basal temporal lobe, posterior cingulate/precuneus, lateral temporal-parietal neocortex, and subfrontal regions. Figures 1 and 2 are maps of GM loss in the AD test and AD retest groups, respectively (p⬍0.05, FDR corrected). Little difference is seen in the topography of rate of GM loss between the completely independent test and retest groups of AD subjects. A test vs. retest difference map showed no significant voxels, thus illustrating the stability of the TBM method. Conclusions: We implemented an automated TBM algorithm in SPM5. The algorithm demonstrated excellent test-retest reproducibility in independent subject samples. The maps displaying rates of atrophy agree quite well with the expected anatomic distribution of greatest neuronal pathology in AD.