اقدام نظارت بر اختلال در ناتوانی در ادراک بیماری برای همی پلژی

Abstract Every movement begins with action programming, and ends with a produced effect. Anosognosia for hemiplegia (AH), involving unawareness of motor deficits after brain damage, is a striking but also poorly understood symptom in clinical neurology. It has been suggested that it may result from a combination of cognitive and sensorimotor dysfunctions, including impairments in monitoring motor action and detecting the mismatch between intention and outcome. Here we investigated the relationship between motor action awareness and monitoring of self-produced movements by using a motor imaginary task, which was performed with either the intact or the affected limb. We tested 10 right brain-damaged patients, including 5 with AH, in comparison with 5 healthy controls. In a first phase, participants were asked to either realize or imagine a movement with their right or left arm. In a subsequent recognition phase, the participants had to recall whether the movement was a realized or imagined and which arm was used. AH patients performed significantly worse relative to no-AH patients and healthy controls for the left movements. Specifically, we found that AH patients believed they had realized movements with their (paralyzed) left arm even when they failed in the left execution condition. However, they also made more errors for movements actually realized with the right hand. These findings confirm that impaired action monitoring may contribute to AHP. Furthermore, our results support the notion of an action control system integrating “feedforward” signals through a comparison process between the intention and execution of movement, but also indicate that monitoring deficits in AHP are not strictly unilateral. Combined together, dysfunction of motor comparator processes and more general monitoring deficits may add up to lead to unawareness of paralysis.

Introduction Babinski (1914; see Langer & Levine, 2014) described for the first time anosognosia as a neurological disorder that is characterized by a lack of interest and/or ignorance of brain-damaged patients for their deficits, such as hemiplegia after stroke (the most common presentation of anosognosia). These patients fail to recognize a severe motor loss despite direct confrontation during the neurological examination or everyday activities. This unawareness of paralysis may manifest itself in inappropriate behaviors or incorrect subjective reports (e.g., the patient claims that he/she is able to move the paralyzed left arm or that he/she has just moved it), but also in unrealistic judgments about the consequences of the deficit for actions (e.g., writing, dressing) and/or unrealistic decisions (e.g., the patients may want to get out of bed or go home). Hence, anosognosia for hemiplegia (AHP) has obviously important implications for clinical management and rehabilitation (Jenkinson, Preston, & Ellis, 2011). It is generally associated with poorer recovery and greater burden for caregivers. However, the cognitive and neural mechanisms of these phenomena are still poorly understood. Moreover, they are difficult to study systematically because of the prevalence of symptoms in the acute phase and the heterogeneity of clinical manifestations (Vocat, Staub, Stroppini, & Vuilleumier, 2010). Many theories have been proposed since the initial description of this syndrome by Babinski (1994), but none appears definitive or sufficient to explain all aspects and variations of anosognosia (Jenkinson et al., 2011, Orfei et al., 2007 and Vocat and Vuilleumier, 2010). Babinski and others (Langer & Levine, 2014) have insisted on the role of proprioceptive deficits, but although these are frequent and contribute to AHP, double dissociations are commonly observed (Cocchini et al., 2010 and Vocat et al., 2010). The same applies to spatial neglect including personal and body-centered forms (Vocat et al., 2010). Several scholars including Heilman and colleagues (Gold et al., 1994 and Heilman et al., 1998) as well as Berti and colleagues (Pia et al., 2004 and Spinazzola et al., 2008) highlighted the role of deficits altering brain mechanisms for the control of movement, notably based on current theories of movement control in the normal state (Frith, Blakemore, & Wolpert, 2000). According to these viewpoints, unawareness of motor paralysis might result from a loss of motor intention (Heilman, 2014) or a destruction of monitoring processes that normally allow the brain to compare the actual motor outputs with motor commands (Wolpert, Ghahramani, & Jordan, 1995). Such deficits may lead to an inability to detect a failure in moving the affected limbs. This view accords with current theories of motor control and motor awareness (Wolpert, et al., 1995) and helps explain a number of intriguing clinical phenomena (Jenkinson & Fotopoulou, 2010). For instance, patients with AHP appear unable to distinguish between the “real” production of movements and purely internal representations of movements (Fotopoulou et al., 2008, Jenkinson et al., 2009 and Pia et al., 2013), Moreover, motor imagery can still operate for the paralyzed limb despite AHP (Garbani et al., 2012; for review Vuilleumier, 2000). In parallel, however, there is evidence that AHP is not a unitary phenomenon but varies according to the test procedures or patients. For instance, Marcel, Tegner, and Nimmo-Smith (2004) showed that judgment of motor performance is particularly overestimated by patients with AHP in self-related conditions (first person perspective), rather than when the same judgments are made by imagining another person in the same conditions (third-person perspective), which may reflect a more general weakening overall cognitive functioning or more extensive monitoring deficits. In addition, patients with AHP may show deficits in monitoring processes not directly related to movement but related to perception of visual events (Feinberg, 2007 and Jenkinson et al., 2010) or to performance in other cognitive tasks (Marcel et al., 2004 and Vocat et al., 2013), which might also contribute to the pathogenesis of AHP. Given the difficulty to pinpoint a single mechanism responsible for AHP, several recent hypotheses (e.g., Davies et al., 2005, Fotopoulou, 2012, Fotopoulous, 2014, Moro, 2013 and Orfei et al., 2007; Prigatano, 2009; Starkstein et al., 2010 and Vuilleumier, 2004) have proposed that this phenomenon may emerge from a combination of deficits (see Vocat & Vuilleumier, 2010). For instance, Davies et al. (2005) suggested that AHP results from both an illusion (leading to delusional beliefs) and impaired evaluation of the belief process (leading to uncritical acceptance of the illusion). Similarly, Vuilleumier (2004) proposed a general “ABC model” of anosognosia, where a combination of deficits affecting at least three main domains (assessment, belief, and control operations) may interact in order to give rise to awareness of motor losses, or instead cause anosognosia or denial (see also Vocat and Vuilleumier, 2010 and Vuilleumier et al., 2013). Hence, besides impairment in motor control (Bottini et al., 2009, Heilman, 2014 and Pia et al., 2004), additional deficits in cognitive and motivational/emotional processes related to performance monitoring and belief shifting could also contribute to anosognosia (i.e., Check operations in ABC model). This hypothesis is supported by recent findings in a “riddle test” (Vocat et al., 2013). In this task, patients have to find a target word, initially unknown to them, based on successive verbal cues, and indicate their level of confidence for each response. Patients with AHP were found to exhibit abnormally high confidence in their guesses, unlike hemiplegic patients without AHP, reflecting selective difficulties to appraise uncertainty and to experience doubt about uncertain beliefs, even when contradicted by external evidence (e.g., additional riddle cues incompatible with initial guess). In a detailed case study, Venneri and Shanks (2004) also showed that persistent anosognosia is associated with a failure to assess the veracity of mental contents about current abilities beyond motor control. More recently, Fotopoulou (2012) proposed a more general model based on computations of ‘prior beliefs’ and ‘prediction errors’. According to this view, the ability to change beliefs and the ability to monitor performance under uncertainty could play a critical role in AHP in combination with other primary deficits in sensory, motor and/or attentional functions. In line with the heterogeneity and complexity of clinical manifestations of AHP, its neural substrates also remain unresolved. Following the pioneer work of Bisiach (Bisiach & Geminiani, 1991) and Heilman (1991), anosognosia has generally been attributed to lesions in the right parietal lobe. More recently, two studies (Berti et al., 2005 and Gandola et al., 2014) highlighted a high prevalence of lesions in right frontal and premotor cortical regions. However, other findings (Karnath & Baier, 2010) point to critical involvement of the right insula, particularly its rear part, together with adjacent white matter. Moreover, a recent study (Vocat et al., 2010) showed that different brain regions might contribute to different aspects of AHP at different stages after stroke. In hyper-acute stages, the severity of AHP correlates mainly with lesions in insula, anterior part of the internal capsule, basal ganglia and para-ventricular white matter of the right hemisphere; whereas for longer lasting AHP in post-acute phase, the severity of anosognosia correlates with additional involvement of the premotor cortex, anterior cingulate gyrus, temporo-parietal junction, and medial temporal regions (hippocampus and amygdale). This anatomical diversity highlights the role of multiple components in AHP. In the present study, we specifically tested whether self-monitoring deficits for movements in patients with AHP are selective for the affected limb, as would be expected if it is due to selective damage to a motor comparator process controlling motor action (Frith et al., 2000 and Wolpert et al., 1995), or whether instead the monitoring deficit would extend to the non-paralyzed limb, as would be expected if AHP is at least in part due to more general deficits in reality checking operations (Vuilleumier, 2004 and Venneri and Shanks, 2004). To this aim, we built on the elegant work of Jenkinson et al. (2009) who examined the ability of patients to discriminate between motor information generated internally (by intention) or externally (by execution and proprioceptive experience). Their paradigm used an approach previously developed to test reality monitoring capabilities in psychiatric disorders where awareness of motor control is also impaired (e.g., schizophrenia where delusions of alien movements are common; see Subramaniam et al., 2012). In the paradigm designed by Jenkinson et al. (2009), monitoring capacities were estimated in a memory task for action information (gestures) and visual information (drawings). In the visual domain (real or imagined drawings), AHP patients showed greater difficulty in distinguishing between these two types of memory (real or imagined) relative to stroke patients without AHP or to healthy controls (Jenkinson et al., 2009). In the motor domain (gestures executed, observed, or imagined), hemiplegic patients with and without AHP were similarly impaired at distinguishing between the different memories, in contrast to the healthy controls. These results appear consistent with the idea that deficits in reality monitoring associated with anosognosia may extend beyond the motor domain, because specific deficits for self-made movements were not only seen in anosognosic patients but also in those with motor loss without anosognosia. These data therefore suggest that dysfunctions in comparator processes or check operations in AHP may not be restricted to the processing of motor information only. However, in this study, patients with AHP also had more severe spatial neglect than those without AHP. It remains possible that neglect and anosognosia for visuo-spatial disturbances (Berti et al., 1996 and Vocat et al., 2010) might have partly affected their performance on the visual reality monitoring task. Furthermore, it is unknown whether similar motor monitoring deficits would be observed for the non-paralyzed arm in patients with AHP, in support of a more global non-lateralized monitoring deficit. In our study, we therefore specifically compared reality monitoring for executed and imagined movements with both the left (affected) and right (non-affected) limb, in patients with and without AHP. If anosognosics primarily rely on beliefs about their health and motor abilities, but are unable to verify their beliefs according to the real situation despite normal motor commands and normal sensory feedback, they might show abnormal reality monitoring even for actions performed or imagined in non-neglected space. However, if AHP is crucially dependent on motor control processes, monitoring deficits should predominate for the affected limb. Surprisingly, no prior study has hitherto compared both limbs in similar tasks.

Results 3.1. Neuropsychological results We also examined performance on several standard neurological and neuropsychological tests (Table 1). Compared to HP, patients with AHP showed no significant difference in any of these tests, including tactile sensation (Mann–Whitney U-test; U = 6, p = .1548), motor (U = 7.50, p = .133), visual field loss (U = 10, p = .317), visual extinction (U = 10, p = .512), spatio-temporal orientation (correct in all patients), as well as visuo-spatial neglect tasks (U = 6, p = .179). Most importantly, no difference was found for tests probing general memory function, including digit span (U = 10, p = .601), verbal memory (U = 10, p = .512), and working memory (U = 4, p = .075). MMSE scores were also comparable in AHP and HP groups (26.5 and 27.5, respectively, U = 10, p = .832). 3.2. Action monitoring results Performance in the action monitoring task was assessed by examining three different indices of recall for each participant: number of ‘correct responses’ in each action condition; number of answers from each ‘category of response’ (regardless of accuracy); and number of ‘specific errors types’ in specific conditions (i.e., misattribution responses Left → Right or Im → Real; cf. Table 2). 3.3. Group results The number of ‘correct responses’ in each condition was averaged across participants for each group and then compared between groups using a Kruskal–Wallis one-way analysis of variance by ranks. This global analysis demonstrated a significant difference among the AHP, HP and CTRL groups for several conditions: Left/Realized (KW = 14.00; d.f. = 2; p < .001), Left/Attempted (KW = 13.42; d.f. = 2; p = .001), Left/Imagined (KW = 11.74; d.f. = 2; p = .002), but also Right/Realized (KW = 11.02; d.f. = 2; p = .004), Right/Imagined (KW = 9.76; d.f. = 2; p = .007), and New (KW = 10.27; d.f. = 2; p = .005) actions. Subsequent pairwise comparisons between groups revealed significant differences between the AHP and HP patients (p < .01) concerning the Left/Attempted, Left/Imagined and New actions, but no difference for other conditions. Relative to CTRL, the AHP group differed a significant differed for all conditions (p < .01) except the Lef/Attempted actions; whereas the HP group differed from CRTL only for Left/Realizedcondition (p = .024). Thus, overall, the HP group showed a pattern of performance globally identical to the CTRL group except for the more frequent (correct) recall of motor failures. Thus, they could generally report that they had attempted to move their paralyzed limb (even if no movement occurred) and made no confusions with imagined movements. Moreover, the number of actions in the Left-Realized condition correctly recalled as “attempted” by HP patients (due to their left paralysis) was similar to the number of actions correctly recalled as “executed” by CTRL for the same condition ( Fig. 2). By contrast, the AHP patients produced a high number of incorrect responses to the Left-Realized actions compared to both the CTRL and HP patients. Strikingly, AHP patients usually reported having executed these actions despite their failure to move the left limb. Unlike the HP group, they never reported “Attempted” movements. Furthermore, they made similar errors in the Left-Imagined condition, i.e., these actions were most often recalled as being executed (see Fig. 2). We next computed the number of answers for each ‘category of responses’ given in the recall phase (Left/Realized, Left/Imagined, Right/Realized, Right/Imagined, New, Attempted) given across all conditions in each group, averaged across participants (see Fig. 3). Again, a Kruskal–Wallis one-way analysis of variance by ranks showed a significant difference among the AHP, HP, and CTRL groups for all conditions (N = 15): Left/Realized (KW = 13.10; d.f. = 2; p = .001), Left/Imagined (KW = 12.14; d.f. = 2; p = .002), Right/Realized (KW = 9.78; d.f. = 2; p = .007), Attempted (KW = 13.42; d.f. = 2; p = .001), and New responses (KW = 10.27; d.f. = 2; p = .005), but not for Right/Imagined (KW = 2.80; d.f. = 2; p = .246). Follow-up comparisons between pairs of groups indicated a difference between the AHP and the HP groups (p < .02) concerning the frequency of Left/Realized, Left/Imagined, Attempted, and New responses, but no difference for other comparisons. Relative to CTRL, the AHP showed significant differences for all responses (p < .01) except the Left/Realized, Right/Imagined, and Attempted. In contrast, the HP group only differed from the CRTL group for Attempted response categories (p = .025). Thus, overall, AHP was associated with poor discrimination in recall between realized and imagined movements with the left/paralyzed limb. Further, AHP was associated with a general and paradoxical increase for making responses from the “left realized” category ( Fig. 3), despite their left paralysis. In other words, incorrect recall of (non-executed) left-sided movements was the most common error made by patients with AHP, regardless of the original encoding condition. Finally, the analysis of the distribution of 'specific error types' showed that the AHP group produced a higher number of incorrect recall responses across all task conditions (KW = 10.62; d.f. = 2; p = .004), relative to the other two groups (AHP vs HP: p = .04; AHP vs CTRL p = .004). However, across conditions, AHP patients produced the same total number of recall responses as the HP and CTRL, suggesting an absence of global memory loss in the AHP group but rather a confusion between the different execution conditions. Indeed, omission errors (i.e., responding “new” to an old instructed action in the recall phase, Fig. 4C) did not significantly (KW = 1.08; d.f. = 2; p = .581) differ between the three groups overall (all pairwise comparisons p > .10). Critically, the AHP group showed a selective divergence from the others for several misattribution errors, i.e., Im → Real for both hands (KW = 11.74; d.f. = 2; p = .002; AHP vs HP: p = .049; AHP vs CTRL p = .007), Right → Left (KW = 11.78; d.f. = 2; p = .002; AHP vs HP: p = .048, AHP vs CTRL p = .015), and Left → Right (KW = 7.67; d.f. = 2; p = .021; AHP vs HP: p = .24; AHP vs CTRL p = .008). For the misattribution Real → Im (inversion errors, KW = 9.32; d.f. = 2; p = .009), only the CTRL group differed from AHP (p = .012), whereas the difference between AHP and HP was not different (p = .412). Thus, the most common type of misattribution in AHP concerned the imagined actions that tended to be reported as having been realized ( Fig. 4A). When inspecting misattributions for each hand separately, only the misattribution Im → Real for the Left hand (Fig. 4B) were significantly higher in the AHP group compared to the other two (KW = 11.74; d.f. = 2; p = .002; AHP vs HP: p = .048; AHP vs CTRL p = .007). This reflects a characteristic pattern of anosognosia errors. However, in addition, the Im → Real errors for the Right hand were also significantly (KW = 7.28; d.f. = 2; p = .026; AHP vs HP: p = .034; AHP vs CTRL p = .010). For inversion errors, involving the incorrect recall of a new (non-executed) action, the AHP group also differed from both other groups (Fig. 4C; KW = 12.02; d.f. = 2; p = .002; AHP vs HP: p = .049; AHP vs CTRL: p = .001). Moreover, these news actions were most often attributed to an action realized with the left hand in the AHP group. Finally, errors were generally more frequent for the left hand than the right hand in the AHP group. Errors did not differ between the two sides in other groups (p = .347). 3.4. Individual results Our group analysis was completed by further inspection of single-case data to verify the consistency of impairments across individual patients. Table 4 shows the same analyses as above but now performed for each single patient with AHP relative to the average number of answers given for each ‘category of responses’ in the other groups (HP and CTRL). For the Left/Realized responses, every single AHP patients showed a reliable difference with respect to the other groups. For the Left/Imaged responses, all 5 patients showed significant differences in comparison with the CTRL group and 3 in comparison with the HP group (but with a similar tendency for 2 remaining patients, p = .056). The attempted response also showed significant difference between every single AHP patient and the HP patients ( Table 4). For other response categories, only one AHP patient differed from other groups ( Table 4). Table 4. Analysis of each AHP patient compared to others groups for the ‘type of response’, ‘specific error types’ and ‘correct response’ using the Bayesian hypothesis test (two-tailed probability) according to Crawford statistical analysis. Type of response AHP_1 AHP_2 AHP_3 AHP_4 AHP_5 Left/Realized AHP vs HP <.001 <.001 <.001 <.001 <.001 AHP vs HC <.001 <.001 <.001 <.001 <.001 Left/Imaged AHP vs HP .021 .056 .056 .021 .009 AHP vs HC <.001 <.001 <.001 <.001 <.001 Right/Realized AHP vs HP .268 .018 .268 .268 .268 AHP vs HC .056 .003 .056 .056 .056 Right/Imaged AHP vs HP .706 .071 .706 .706 .706 AHP vs HC .95 <.001 .95 .95 .95 Novel AHP vs HP .080 .080 .080 .001 .004 AHP vs HC .080 .080 .080 .001 .004 Attempted AHP vs HP .010 .010 .010 .010 .010 AHP vs HC – – – – – Correct response AHP_1 AHP_2 AHP_3 AHP_4 AHP_5 Left/Realized (Executed) AHP vs HP – – – – – AHP vs HC <.001 <.001 <.001 <.001 <.001 Left/Realized (Attempted) AHP vs HP – – – – – AHP vs HC .010 .010 .010 .010 .010 Left/Imaged AHP vs HP .025 .010 .025 .010 .010 AHP vs HC <.001 <.001 <.001 <.001 <.001 Right/Realized AHP vs HP .080 .004 .080 .080 .543 AHP vs HC <.001 <.001 <.001 <.001 <.001 Right/Imaged AHP vs HP .224 .025 .224 .224 .703 AHP vs HC <.001 <.001 <.001 <.001 <.001 Novel AHP vs HP .080 .080 .080 <.001 <.001 AHP vs HC .080 .080 .080 <.001 <.001 Type of error AHP_1 AHP_2 AHP_3 AHP_4 AHP_5 Anosognosia error (ImLeft − <RealLeft) AHP vs HP .011 .003 .003 .003 .001 AHP vs HC <.001 <.001 <.001 <.001 <.001 Inversion error AHP vs HP .242 .256 .242 .230 .157 AHP vs HC .072 .074 .072 .071 .063 Memory error AHP vs HP .699 .699 .699 .921 .699 AHP vs HC – – – <.001 – Table options The single-case analysis for ‘correct responses’ showed a similar profile with a significant differences in all cases when the left hand was involved (p < .001). With the right hand, only one AHP patient showed a difference with HP ( Table 4) but all AHP patients were significantly different than the HC group (p < .001). Finally, the critical anosognosia errors (Table 4) were more frequent in each and every AHP patient compared to other participants, as verified by our single-case analyses. For anosognosia errors, all 5 patients showed significant differences (Crawford test) relative to both the healthy control group (p < .001) and the HP group (p < .001). This analysis thus confirms that patients with AHP showed a distinctive tendency to report that “imagined” movements were “realized”. This tendency was seen in all cases. In keeping with the group results above, CTRL and HP patients did not differ ( Table 4). 3.5. Lesions results Finally, we performed an exploratory anatomical analysis comparing the lesions overlap of AHP and HP patients. Using a standard subtraction analysis (Fig. 5), the maximal lesion overlap across all patients showed a relative predominance of damage to the temporo-parietal junction and insula in AHP as compared with the HP group (MNI coordinates: 36, 20, 7). This accords with previous findings (Baier and Karnath, 2005 and Vocat et al., 2010) indicating that these cortical areas are frequently associated with a greater severity of anosognosia. An additional statistical comparison was also performed with a voxel-by-voxel statistical lesion mapping (VLSM) procedure with the rate of anosognosia errors and inversion errors. This analysis showed that the most significant peak of lesion differences arose in the posterior insula (MNI coordinates: 41, 5, 9) for anosognosia errors (Fig. 6), but in the anterior insula (MNI: −51, 23, −3) and white-matter underneath the inferior frontal gyrus (MNI: 37, 20, 24) for inversion errors (Fig. 6). However, these anatomical results must be taken as essentially descriptive given the small patient sample overall. Voxelwise statistical anatomical analysis comparing lesions of anosognosic and ... Fig. 6. Voxelwise statistical anatomical analysis comparing lesions of anosognosic and nosognosic patients using a VLSM analysis for anosognosia (ImLeft → RealLeft) and inversion (New → Real or Im) errors. IPG: inferior parietal gyrus; STG: superior temporal gyrus; MTG: middle temporal gyrus; PSA: primary somatosensory area; MCC: Middle Cingulate Cortex; IFG: Inferior Frontal Gyrus; OFC: Orbito-Frontal Cortex.