INVESTIGADORES
BELZUNCE MartÍn Alberto
congresos y reuniones científicas
Título:
[18F]FDG PET Templates and metabolic connectivity matrices for Alzheimer's disease research
Autor/es:
SOL CATALDO; MARTIN A. BELZUNCE
Lugar:
Buenos Aires
Reunión:
Congreso; XXIX Congreso de la Asociación Latinoamericana de Sociedades de Biología y Medicina Nuclear (ALASBIMN); 2023
Resumen:
ObjectivesTheobjective of this work is to create [18F]FDGPET templates and intersubject metabolic connectivity matrices ofcognitive normal (CN) and Alzheimer’s disease (AD) populations thatcan be used as tools in Alzheimer's disease research.Materialsand methods Weobtained a set of 70 [18F]FDGPET and MRI images from the Alzheimer’s Disease NeuroimagingInitiative (ADNI) database. The PET images consisted of 6 frames of 5min each and the MRI were high-resolution T1-MPRAGE anatomicalimages. The PET image was registered to the MRI and both images werenormalized to the MNI152 space. Then, the acquisition values of thePET image were normalized to the cerebellum uptake value to producean intensity normalized image. Regional uptake values were computedusing the 95 regions defined in the Hammers’ atlas. Outliers foreach group were defined as images with at least 10 regional uptakevalues greater than three scaled median absolute deviations (MAD)from the median. To createPET templates, the intensity-normalized images were averaged for eachgroup excluding the outliers. Toestimate the intersubject metabolic connectivity, the 95 regionaluptake values per subject were concatenated creating a matrix withsizes of 95×Ns,being Nsthe number of subjects for the CN and AD groups. Pearson'scorrelation coefficients were then estimated between the uptakevalues from each pair of regions. Finally, the templates were used toevaluate a simple AD classifier based on [18F]FDGPET images. The first method consisted in classifying each subjectwith the class of the closest template using the Frobenius distancebetween the subject image and each template as a metric. Aleave-on-out scheme was used to evaluate the accuracy of theclassifier.ResultsSeventyimages were processed successfully. Four and two outliers weredetected for the CN and AD groups respectively. The [18F]FDGPET templates (Figure 1-A) and the intersubject connectivity matrices(Figure 1-B) were successfully generated for each group using datafrom 32 subjects for each of them. The [18F]FDGPET AD template shows a general pattern of hypometabolism and is moreprofound in the posterior temporal lobe as expected. The connectivitymatrices show significant differences that may be useful inpredicting Alzheimer's disease with FDG. The AD classifier based onthe Frobenius distance to each atlas had an accuracy of 78%,correctly predicting 26 of the 34 AD cases, and 29 of the 36 CNcases.ConclusionsThe[18F]FDGPET templates can be used in the study of neurodegenerative diseasesand in the development of new AD biomarkers and predictors, as provedin this work with a simple example. The intersubject metabolicconnectivity demonstrates promising outcomes and has the potential toevolve into a novel biomarker for AD.p { line-height: 115%; text-align: left; orphans: 2; widows: 2; margin-bottom: 0.1in; direction: ltr; background: transparent }a:link { color: #000080; text-decoration: underline }a:visited { color: #800000; text-decoration: underline }