Profiling cognition and brain metabolism in amyotrophic lateral sclerosis and frontotemporal dementia

: Amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD) are described as a disease continuum, given their shared clinical, genetic, and pathological characteristics. The comparisons of clinical and biomarker features within ALS and behavioral variant FTD (bvFTD) spectrum, would be of utmost importance for diagnostic and prognostic purposes. This study investigated biomarker differences between patients with ALS cognitively-normal (ALS-cn), ALS-FTD, and bvFTD. Participants, genetically screened for known ALS- and FTD-associated mutations, underwent neuropsychological assessments, CSF analysis, and brain imaging through 18-fluorodeoxyglucose PET ([18F]FDG-PET). Neuropsychological data were analyzed by calculating, for each cognitive domain, a composite score by averaging the rank-transformed z-scores of all tests measuring the same domain. [18F]FDG-PET analysis was performed using a validated voxel-based SPM method at single-subject and group-level. To evaluate the ability of the identified markers to differentiate ALS-cn, ALS-FTD, and bvFTD, machine-learning models-including support vector machine (SVM) and random forest (RF)-were applied, offering a streamlined, data-driven approach to improve diagnostic precision across this spectrum of disorders. 20 ALS-cn, 19 ALS-FTD, and 21 bvFTD patients were included. Neuropsychological composite z-scores revealed significant differences across groups, underlining worse performance in bvFTD regarding memory, visuospatial, language and executive functions. Brain [18F]FDG-PET showed a pattern of hypometabolism increasing from ALS-cn to ALS-FTD and reaching its greatest extent in bvFTD. Specifically, brain hypometabolism was mainly confined to the sensorimotor cortices and the frontobasal regions in the ALS-cn group, whereas in the ALS-FTD group it was extended to the supplementary motor area and the dorsolateral frontal cortex, and in the bvFTD group, a widespread hypometabolism further affected the frontomesial and orbitofrontal cortices. No significant differences in CSF biomarkers were observed. SVM correctly classified 83% of patients, indicating a good level of classification performance, while RF showed perfect accuracy (100%). The two models shared eight to ten most relevant features in the classification system, namely age, disease duration from symptoms onset to diagnosis, total composite z-score, superior frontal gyrus (left), middle frontal gyrus (left), middle frontal gyrus - pars orbitalis (left and right), and anterior cingulate cortex (left). Our study identified significant differences in the biomarkers according to the neurodegenerative clinical groups within the same disease spectrum. These differences were evident in neuropsychological profiles and brain hypometabolism patterns, successfully addressing the study's aim and providing valuable insights for differential diagnosis into ALS-FTD continuum heterogeneity.