Atrophy Overlap Map PCA on Lesion maps: Brain components that correlate with language components Deconstructing the transdiagnostic nature of language symptoms in frontotemporal dementias Lorna Garcia-Penton 1,2 , Ajay D. Halai 1 , Siddharth Ramanan 1 , Nikil Patel 3 , Ruth U. Ingram 4 , Stefano F. Cappa 5,6 , Karalyn Patterson 1,2 , James B. Rowe 1,2 , Peter Garrard 3 , Matthew A. Lambon Ralph 1 1 MRC-CBU & 2 Dept. Clinical Neurosciences, University of Cambridge, UK; 3 St. George’s, University of London, UK; 4 University of Manchester, UK; 5 IUSS & 6 IRCCS Pavia, Italy Participants: Patients and healthy controls (HC) Neuroimaging: T1-MRI, DW-MRI. Analyses: GM segmentations and FA maps were calculated and used to estimate the degree of abnormality for each brain voxel (fuzzy lesion map). a) Principal component analysis (PCA) of fuzzy lesion maps and MLSE metrics to identify covarying patterns of language and brain abnormalities. b) Voxel-based correlation mapping of emergent language dimensions (GM/FA maps combined). c) Multiple regression analysis with regions of interest (ROI) to pinpoint brain-language associations. Methods GM/FA intensity and PCA language performance: PC1:Motor-Speech /Phonology PC2: Semantic PC3: Syntax Voxel-based correlation mapping We thank all the patients, their families and carers for their support. This study was funded by MSCA-IF (no. 893329 to LGP), The Rosetrees Trust (no. A1699 to ADH and MALR), ERC (GAP: 670428 to MALR), the MRC (MR/R023883/1 to MALR; MR/V031481/1 to ADH) Acknowledgements • Frontotemporal lobar degeneration (FTLD) related syndromes featuring primary language impairments are known as primary progressive aphasia (PPA). • PPA is classified into three variants: semantic (sv), non-fluent (nfv), and logopenic (lv). Yet their language profiles overlap, and some patients do not fit any variant (i.e., 'mixed' PPA). • Language deficits also can occur across other FTLD-related syndromes like progressive supranuclear palsy (PSP) and corticobasal syndrome (CBS), suggesting that language symptoms are transdiagnostic. • Note that lvPPA is usually associated with Alzheimer's disease (AD) pathology, even if it falls under clinical FTD-PPA syndromes. Similarly, CBS can be underpinned by AD. • Though brain atrophy determines PPA language deficits, similar lesion profiles often have different language deficits, while different lesions can cause similar impairments (e.g. lvPPA / nfvPPA). • We apply a multimodal approach to behavioural and neuroimaging data (cf. Ingram et al 2020, Brain; Murley et al 2020, Brain): To link language dimensions of variations across FTD-PPA spectrum to their underlying neuroanatomical factors. Introduction Varimax rotated factor loading: Varimax rotated factor scores: Language symptoms dimensions • As expected PCA found three language dimensions—motor-speech/phonology, semantics, and syntax—and thirteen brain atrophic dimensions, including temporal, frontal, and parietal regions. • Semantic linked strongly with temporal lobe atrophy, but other atrophic and language dimensions did not, suggesting a non-one-to-one neural mapping of language functions. • Multiple regression showed that atrophy in a distributed set of frontoparietal regions predicts motor-speech, phonology, and syntax deficits, while atrophy in ATL predicts semantic deficits. Atrophy and language variations on across FTD- PPA syndromes are deconstructed on a brain- behaviour continuum, with only semantic variants being clearly separate. lg620@cam.ac.uk ccpp.cam.ac.uk ftd.neurology.cam.ac.uk @CambridgeFTD Demographic Mean (std) Variables svPPA nfvPPA lvPPA CBS PSP HC Age 66.3 (1.8) 71.2 (1.6) 70.2 (1.4) 70 (1.8) 68 (2.0) 65.6 (1.1) Education (y) 18.9 (0.8) 17.3 (0.7) 19.2 (0.6) 18.3 (0.8) 17.1 (0.9) 21.1 (0.5) Female/Male 7/6 10/5 6/16 6/7 4/6 13/19 Sample size 13 15 21 13 10 32 Behaviour: Participants completed the Mini-linguistic State Examination (Patel et al 2021, Brain Communications). Brain atrophic dimensions Discussion 15 bilateral GM ROIs to predict PCA language performance (stepwise regression): A) Motor-Speech/Phonology B) Semantic C) Syntax ROI-based multiple regression Left Inferior Frontal Gyrus / Precentral [-44, 12, 32] MLSE-PC3 scores Left anterior Inferior Temporal Gyrus [-56, -6, -382] MLSE-PC2 scores Left Rolandic Operculum [-52, 4, 12] MLSE-PC1 scores Correlation Peak p < 0.05 FWE cluster-level corrected Sentences repetition Non-word reading Word reading Sentence comprehension Semantic association Non-word repetition Pointing Repetition Apraxia Naming Writing Picture description Z-scored, 3 PC retained, cum. exp. var. 66.74% MLSE Variables Motor-Speech/Phonology Semantic Syntax Loadings PC1: Motor-Speech/Phonology (EV = 41.44 %) PC2: Semantic (EV = 16.59 %) PC3: Syntax (EV = 8.71 %) PC2: Semantic PC3: Syntax PC1: Motor-Speech/Phonology Z-> 3 4 6 8 10 12 MLSE-PC1 (Motor-speech/Phonology) Brain-PC9 scores Varimax rotated scores svPPA nfvPPA lvPPA CBS PSP PC9: R pSTS / SMG / Rolandic Operculum Brain-PC6 scores r = -0.60 p = 0.000 MLSE-PC2 (Semantic) Varimax rotated scores svPPA nfvPPA lvPPA CBS PSP PC6: R Temporal Pole (TP) / Insula (INS) MLSE-PC2 (Semantic) Brain-PC1 scores r = -0.61 p = 0.000 Varimax rotated scores svPPA nfvPPA lvPPA CBS PSP PC1: L ant & mid Temporal Lobe (ATL) Varimax rotated scores svPPA nfvPPA lvPPA CBS PSP MLSE-PC2 (Semantic) Brain-PC8 scores r = -0.70 p = 0.000 PC8: L Temporal Pole (TP) / Fusiform (FF) Brain-PC3 scores MLSE-PC2 (Semantic) Varimax rotated scores svPPA nfvPPA lvPPA CBS PSP PC3: Caudate / Putamen r = -0.35 p = 0.004 r = -0.35 p = 0.003 AAL 3 predictors, F (3,66) = 6.09, p = .016, R 2 = 0.29 Rolandic Operculum Mid Temporal Pole Precentral Gyrus MLSE-PC1 scores 3 predictors, (F (3,66) = 6.09, p = .016, R 2 = 0.29) MLSE-PC3 scores IFG pars triangularis Mid Temporal Pole Angular Anterior Inferior Temporal Gyrus MLSE-PC2 scores 1 predictors, F (1,68) = 88.36, p < .001, R 2 = 0.57