Comparing the Value of Dynamic vs. Static-Image-Based Tests of Emotion Recognition in Neurodegenerative Diseases

Abstract

Introduction: More precise subtyping within dementia syndromes leads to better prediction of pathology, supporting individualized, disease-specific treatments. Notably, studies highlight that identification of the right-temporal or semantic behavioral variant frontotemporal dementia (sbvFTD) subtype relies in part on measuring emotion recognition abilities. Methods: To evaluate the effectiveness of current tools, we compared two dynamic video-based affect labeling tests - the Dynamic Affect Recognition Test (DART) and The Awareness of Social Inference Test-Emotion Evaluation Test (TASIT-EET) - against the static image-based Name Affect subtest of the Comprehensive Affect Testing System (CATS-NA) test. A total of 555 persons with dementia (PwD), in the early stages of neurodegenerative disease (Clinical Dementia Rating <= 1; Mini Mental State Examination > 20), diagnosed with Alzheimer's disease syndrome (AD) (n=154), progressive supranuclear palsy syndrome (PSPS) (n=88), non-fluent variant primary progressive aphasia (nfvPPA) (n=77), semantic variant PPA (n=53), behavioral variant frontotemporal dementia (bvFTD) (n=124), semantic bvFTD (n=65), and 133 healthy older participants underwent emotion testing and structural MRI. Results: All emotion labeling tests differentiated PwD from healthy controls (DART, AUC=0.81; TASIT-EET, AUC=0.84; CATS-NA, AUC=0.72), and FTD with social cognition deficits (sbvFTD, bvFTD, and svPPA) from other PwDs (DART, AUC=0.64; TASIT-EET, AUC=0.66; CATS-NA, AUC=0.63). Dynamic tests outperformed CATS-NA in differentiating sbvFTD from bvFTD and svPPA (DART, AUC=0.79; TASIT-EET, AUC=0.74; CATS-NA, AUC=0.60), whereas DART outperformed TASIT-EET in differentiating sbvFTD from svPPA (DART, AUC=0.73; TASIT-EET, AUC=0.66). Multiple linear regression analysis showed that TASIT-EET performance was predicted by visual memory (Benson-delayed) and verbal semantic (BNT, Animal Fluency) functions (p<0.01) and CATS-NA performance was predicted by visuospatial (CATS-Face matching, Number location) (p<0.001) and executive functions (Modified Trail making speed) (p<0.05), while DART was predicted by only working memory functions (Digit span backward) (p<0.05). DART corresponded to the expected structural anatomy of emotion, including right predominant insula, anterior temporal, and orbitofrontal lobes. While both TASIT-EET and CATS-NA shared that pattern of brain anatomy, TASIT-EET correlated with more left temporal structures than DART, and CATS-NA associated with more dorsal structures than DART. Finally, all emotion labeling tests correlated with real-life empathy deficits measured by a standardized informant-based survey. Conclusion: Tasks showing dynamic audio-visual emotion displays showed better effectiveness for diagnostic differentiation of FTD syndromes than static image-based tasks, and the DART showed better clinical and anatomic precision than the TASIT-EET. Emotion identification deficits are a core feature of dementia syndromes like sbvFTD, but occur in the context of additional cognitive deficits. Therefore, careful selection of tests that reflect the key underlying neural circuits related to emotion, and which minimize demand from other cognitive domains, will result in more accurate diagnoses.