A Pioneering Paper That Provided a Tool for Accurate, Observer-Independent Analysis of ¹⁸F-FDG Brain Scans in Neurodegenerative Dementias (perspective on “A Diagnostic Approach in Alzheimer’s Disease Using Three-Dimensional Stereotactic Surface Projections of Fluorine-18-FDG PET” J Nucl Med. 1995;36:1238–1248)

Minoshima’s innovative paper included in this anniversary issue of JNM built on advances in basic research and clinical brain imaging to present an objective, automated method to assess ¹⁸F-FDG PET scans in individual patients with cognitive decline (1).

With the widespread use of ¹⁸F-FDG in oncology, it is easy to forget that ¹⁸F-FDG evolved from the seminal work of Louis Sokoloff et al. (2) to measure regional cerebral glucose metabolism in small animals with ¹⁴C-deoxyglucose and tissue autoradiography. The method provided a map of local brain function, made possible because regional cerebral glucose metabolism is coupled to neuronal, mainly synaptic, activity.

Pioneering papers published in 1979 extended the Sokoloff method to human brain imaging with PET and ¹⁸F-FDG. These papers were followed in the early 1980s by studies of regional cerebral glucose metabolism and local neuronal activity in neuropsychiatric disease. Scans in Alzheimer disease (AD) showed hypometabolism in the temporoparietal association cortex; this became recognized as the characteristic signature of the disease. At the time, quantitative analysis of brain PET images used regions of interest, comparing radiotracer values in brain regions between patient and control groups. Subsequently, more automated methods such as statistical parametric mapping (3) were developed to detect local changes by mapping PET images onto a stereotactic brain atlas. These methods facilitated comparing groups of PET images on a pixel-by-pixel basis, an approach widely used to analyze ¹⁵O-water images of regional cerebral blood flow in subjects scanned during different neurobehavioral states.

In parallel with the use of PET in brain research was the recognition of the clinical potential of ¹⁸F-FDG images interpreted visually to evaluate patients with possible AD. The work of Minoshima et al. (1), an advance ahead of its time, introduced to the clinic a quantitative automated method of image analysis termed 3-dimensional stereotactic surface projection (3D-SSP). 3D-SSP was objective, accurate, and reproducible, and facilitated the interpretation of PET scans in individual patients suspected of AD.

3D-SSP calculates the statistical significance of differences in cortical metabolic activity between a patient and controls, and projects the data onto a surface rendition of the brain. First, transaxial brain images are spatially transformed to match an anatomically defined, 3D reference brain volume from a stereotactic brain atlas. Second, peak pixel values of cortical metabolic activity values are extracted and projected onto the surface of the brain. Finally, the cortical projection of metabolic data from the patient is statistically compared pixel by pixel to a database of ¹⁸F-FDG scans in age-matched controls to identify pixels beyond the reference range. The results are displayed on the cortical surface as z scores, which indicate the SD of the difference in metabolic value between patient and control data for each pixel.

The paper has several noteworthy features. It demonstrated that for diagnostic purposes, tissue radioactivity measurements worked as well as regional cerebral glucose metabolism values, obviating measurements of arterial radioactivity to implement the Sokoloff model. ¹⁸F-FDG activity values are normalized to the value in a reference region that is minimally or not affected by the disease process, such as the thalamus, pons, or cerebellum. This normalization reduces the variability of regional data due to differences in global metabolism, scanner calibration, and other factors, and increases the sensitivity to detect regional abnormalities. The method focuses on the cerebral cortex, where the characteristic abnormalities of AD are found, reducing the amount of image data considerably and facilitating pattern recognition of abnormalities and interpretation of scans. In fact, data were presented to show that the use of 3D-SSP improved the accuracy of scan interpretation in patients with probable AD.

The 3D-SSP method was applied to other neurodegenerative dementias with a typical pattern of cortical metabolic abnormality, such as dementia with Lewy bodies and frontotemporal dementia, and its ability to differentiate these patterns was demonstrated (4). It has been used in both clinical and research applications and was adapted to SPECT imaging of cerebral blood flow and to PET/MR. Several studies comparing the accuracy of 3D-SSP to standard visual reads showed it to be at least as good as, if not better than, visual reads by expert readers and better than reads by novices. Thus, 3D-SSP helps inexperienced readers, provides confirmation for experienced readers, and can reduce variability among readers. It is a useful educational tool and offers a visual summary of scan results for referring physicians.

In a recent consensus paper supporting the use of ¹⁸F-FDG in neurodegenerative dementia, European associations of nuclear medicine and neurology deemed software tools helpful to assist visual scan reading (5). 3D-SSP software has been used for this task worldwide and is freely available for download (6). The authors’ sharing of their intellectual contribution is both commendable and appropriate, since their work was supported by grants from the U.S. government.

Long before the current interest in artificial intelligence, this pioneering paper provided a tool for accurate, observer-independent analysis of ¹⁸F-FDG brain scans in neurodegenerative dementias, in both the clinical and the research environment. Although published 25 years ago, the concepts it introduced, as well as its practical utility, remain highly relevant.

• © 2020 by the Society of Nuclear Medicine and Molecular Imaging.

• Received for publication June 25, 2020.
• Accepted for publication July 3, 2020.