Anosognosia for Hemiplegia as tripartite disconnection syndrome.

Being aware of one’s own motor performance is crucial for successfully interacting with the world. In rare circumstances, after right brain damage, the delusion of movement can occur despite one’s own motor paralysis, that is the Anosognosia for Hemiplegia (AHP, Babinski 1914). This syndrome offers an extraordinary possibility to explore the neural correlates of motor awareness and understand how people build and maintain their sense of self. Previous neuropsychological and lesional studies pinpointed to the role of action monitoring and error processing failure in AHP due to lesion to lateral premotor cortex and insula. Others hypothesized a cortico-subcortical functional disconnection between top-down, premorbid learned predictions regarding one’s body and the processing of bottom-up "prediction errors" regarding its current state. However, conceptually a damage limited to a single region is insufficient to explain some of the delusional aspects of AHP, such as the inability of patients to learn from falls and social feedback. In fact, the relatively small sample size and the standard methodology of existing Voxel-Based Lesion-Symptom Mapping studies preclude any insights beyond the implication of various, discreet lesion locations in the pathogenesis of AHP (e.g. Berti et al., 2005; Karnath, Bayer, & Nägele, 2005; Vocat, Staub, Stroppini, & Vuilleumier, 2010; Moro et al., 2016). Therefore, for first time, we investigated the hypothesis that motor awareness emerges from a functional system, where separated networks participate in an integrated way (Luria 1980) in the largest cohort of AHP patients to date (n= 95 and 79 hemiplegic controls, disgnosed according to Bisiach test; Biasiach et al, 1986). After the lesion delineation on each patient’s CT and MRI scan, an advanced methodology for lesion analysis (Foulon et al. 2018) allowed us to identify the probability of disconnection of white matter tracts via a posteriori identification of tracts disrupted by the lesion, obtaining maps of disconnections for each patient. Regression analyses were performed on patients’ lesion and disconnection profiles, in order to identify the contribution of grey and white matter structures in AHP, considering groups as independent variable. Differences in age, lesion size, lesion onset- assessment interval and critical motor and neuropsychological deficits were taken into account as nuisance variables. Then, the contribution pattern of the tracts disconnection to AHP (individual or integrated) was tested via Bayesian computation of generalized linear multilevel models. 95 binomial models were computed, starting from the null model (with only the covariates of non-interest) to the full model, with all the covariates of non-interest, the tracts, and all the interactions among them. The regression computed on the lesion location (Figure 1a) shows the involvement of white matter (t = 4.98; p = 0.002), and of grey matter structures previously associated with AHP, such as the insula (anterior long gyrus, t = 4.89; p = 0.002), the temporal pole (t = 4.77; p = 0.003), and the striatum (t = 4.68; p = 0.003). The results from the regression on patients’ disconnection profile (Figure 1b) unveiled a significant contribution of the cingulum (t = 3.85; p = 0.008), the third branch of the superior longitudinal fasciculus (SLF III; t = 4.30; p = 0.003), and connections to the pre-supplementary motor area (preSMA; t = 3.37; p = 0.013), such as the frontal aslant and the fronto-striatal connections. Bayesian statistics confirmed that the disconnection of each tract is critical to AHP (Cingulum, BF10 = 270.98; FST, BF10 = 180.48; FAT, BF10 = 367.61; SLF III, BF10 = 571.49; Figure 2b). However, the model that better fits with our data (99% of probability) includes the contribution of all the four tracts. Our results, derived from the largest lesion mapping study on AHP to date, show that white matter disconnections in three networks contribute to AHP: (1) the limbic network (i.e. connections between the amygdala, the hippocampus and the cingulate gyrus); (2) the ventral attentional network (i.e. connections between temporo-parietal junction and ventral frontal cortex), through the SLF III; and (3) the premotor loop (i.e. connections between the striatum, the preSMA and the inferior frontal gyrus). We demonstrate a tripartite contribution of the three networks to motor awareness. We suggest that bottom up deficits in interoceptive and motor salience monitoring need to be combined with higher-order deficits leading to a multifaceted syndrome in which premorbid beliefs and emotions about the non-paralysed self-dominate current cognition about the paralysed body. We believe that these results will be of broad interest, and will make important contributions to understanding the link between brain function and structure in healthy human and neurological patients.