Cognitive strategies for locomotor navigation in normal development and cerebral palsy

DEVELOPMENTAL MEDICINE & CHILD NEUROLOGY ORIGINAL ARTICLE Cognitive strategies for locomotor navigation in normal development and cerebral palsy VITTORIO BELMONTI 1 SIMONA FIORI 1 ANDREA GUZZETTA 1,2 GIOVANNI CIONI 1,2 ALAIN BERTHOZ 3 1 Department of Developmental Neuroscience, Stella Maris Scientific Institute, Pisa, 2 Department of Clinical and Experimental Medicine, University of Pisa, Pisa, Italy. 3 Laboratoire de Physiologie de la Perception et de l Action, UMR7152 CNRS-College de France, Paris, France. Correspondence to Vittorio Belmonti, Department of Developmental Neuroscience, IRCCS Fondazione Stella Maris, Viale del Tirreno 331, Calambrone, 56128 Pisa, Italy. E-mail: vbelmonti@fsm.unipi.it PUBLICATION DATA Accepted for publication 9th October 2014. Published online ABBREVIATIONS CBT Corsi Block-tapping Task NAI Navigational advantage index Visual spatial impairment is a fundamental disorder in cerebral palsy (CP). However, current spatial testing is restricted to reaching space, whereas navigational space is seldom assessed. The Magic Carpet test, derived from the Corsi Block-tapping Task (CBT) for visual spatial memory, is a new developmental test for navigation. The performances of the Magic Carpet test and CBT were assessed in 17 children with unilateral and bilateral spastic CP. The results were compared with an equal number of typically developing children, matched for age and sex. Magnetic resonance imaging scans of children with CP were scored according to a newly validated semi-quantitative classification. CBT span was significantly lower in CP, especially in bilateral forms, than in the comparison group, whereas the Magic Carpet test span did not significantly differ between the groups. CBT span, but not the Magic Carpet span, was related to gestational age at birth and to basic visual function. Both the CBT span and the Magic Carpet test were related to overall right-hemispheric impairment. In addition, CBT correlated with right periventricular impairment. In CP, navigation is differently impaired than visual spatial memory, and less tightly related to preterm birth, basic visual function, and deep white matter injury. The exploration of navigational space could prove useful in enhancing spatial representation and reference-frame manipulation in CP. Among the accompanying disorders, visual perceptual impairment is particularly relevant in cerebral palsy (CP) (see Ego et al., this issue). The dorsal visual stream, underlying visual spatial and visual motor abilities, seems particularly vulnerable in a variety of early neurological conditions. 1 Strikingly enough, however, psychometric testing is restricted to a narrow partition of the available space: that for reaching and manipulation. Conversely, far, or navigational, space is seldom assessed. Reaching space and navigational space involve quite different cognitive strategies and brain networks. 2 Most importantly, whereas reaching locations are coded relative to the body (egocentric reference frames), more options are available when navigating: either updating egocentric locations or switching to a fully allocentric reference frame. 3 Different brain activation patterns are associated with egocentric and allocentric strategies, the latter not being restricted to the dorsal visual stream, but extending to ventral, medial temporal, and prefrontal areas. 4 The normal development of spatial cognition is still debated: a primarily egocentric model has been recently opposed by more integrative hypotheses. 5 The Corsi Block-tapping Task (CBT 6 ) is a widely used test for spatial memory. CBT performance is linearly related to age in typical development 7 and is impaired in CP. 8 Recently, a computerized navigational version of the CBT, the Magic Carpet test, has been developed from the manual Walking Corsi test, 9 and used to study navigation development. 10 The present study aims to extend the investigation to children with CP. Our hypothesis is that navigation disorders can be dissociated from basic visual spatial deficits, and that different lesions are associated with different impairments. METHOD All patients admitted to IRCCS Fondazione Stella Maris, Pisa, Italy, in 2013, were screened for the following inclusion criteria: age between 5 and 12 years, diagnosis of spastic CP, independent walking without aids, and brain magnetic resonance imaging (MRI) performed after 18 months of age. The final sample was of 17 patients, described in Table I. Basic visual function was scored as 0 (no impairment), 1 (one of mild acuity reduction, unilateral field defect, or strabismus), or 2 (more severe or multiple disorders). Control data were obtained from a sample of 91 typically developing children aged 5 to 12 years, as described elsewhere. 10 From that sample, a subset was extracted by age- and sex-matching them with children with CP, then randomly selecting one matched child per patient. Thus, a larger non-matched comparison group consisting of all 91 typically developing children and a smaller group of 17 2015 The Authors. Developmental Medicine & Child Neurology 2015 Mac Keith Press, 57 (Suppl. 2): 31 36 DOI: 10.1111/dmcn.12685 31

age- and sex-matched children were obtained. The larger non-matched comparison group was retained to provide a double statistical check, because a major part of the variability found in typical development is known to be due to neither age nor sex. 10 The CBT was performed on a plastic board (28cm922.5cm) with nine blocks on top (2.9cm92.9cm93cm). This was part of the BVS-Corsi standardized package. The Magic Carpet is an experimental test, not yet validated psychometrically. It consists of 10 square tiles (30cm930cm), each containing four pressure sensors and a central blue light-emitting diode. Nine tiles are embedded within a grey carpet (310cm9260cm), with the same layout of CBT blocks. The tenth tile, placed outside the carpet, is the starting point. A laptop manages input/output signals. Dark grey panels, 2m high, were placed all around the carpet, except on starting side, so that landmarks were covered. For both tests, the standard short-term memory assessment was applied. On the CBT, the experimenter taps a sequence of blocks, which the patient must retrieve by finger-tapping those blocks in the same order after a start signal. The procedure for the Magic Carpet is identical, except that stimuli are automatically delivered by light-emitting diode switching (tiles are lit up) while the patient stands on starting point, then retrieved by the patient walking on those tiles in the same order. Sequence length identifies the level, increasing from 2 to 9. Five pseudo-random sequences per level are shown, and a level is passed when three of them are correctly retrieved, otherwise the test is terminated. Span is given by the highest level passed, plus an additional 0.33 accredited for every correct response at the highest, non-passed level. This scoring system was used in our previous study on typical development 10 and is aimed at increasing resolution. What this paper adds The Magic Carpet is a new test for assessing locomotor navigation in cerebral palsy. In cerebral palsy, navigation is differently impaired than visual spatial memory in reaching space. The extent of right hemispheric damage correlates with navigation skills and visual spatial memory. CBT and Magic Carpet raw span scores were normalized by age, using the linear regression models fitted on control data. For each test, normalized span was given by span norm =[span (b 0 +b 1 age)]/se, where b 0 and b 1 are the intercept and the slope of controls span-by-age regression line, and SE is the standard error of the residuals. A navigational advantage index (NAI) was similarly obtained by normalizing Magic Carpet by CBT span, on the basis of the linear regression line fitted on control data (in the equation above, substitute span norm with NAI, span with span MC, and age with span CBT ). Raw CBT and Magic Carpet span scores and NAI were compared across groups (CP, matched, and non-matched comparisons) by appropriate t-tests and analyses of variance (ANOVAs) (see Results). MRI was performed using a 1.5tesla Magnetom (GE) scanner. T1-, T2-, and T2*-weighted images were obtained with spin echo, fast spin echo, gradient echo, and fluid-attenuated inversion recovery (FLAIR) sequences on sagittal, axial, and coronal planes or three dimensions. MRI scans were taken in different epochs, but all after the age of 2 years. MRI was not necessarily performed at the same time as behavioural assessment. Structural anomalies were first classified according to Kr ageloh-mann and Horber 11 into brain maldevelopments, periventricular white matter lesions, cortical and deep grey matter lesions, or miscellaneous. Then, a novel semi-quantitative scale 12 was applied (Fig. 1). Anomalies were matched with a Table I: Main characteristics of the sample Patient Age (y) Sex CP form Lesion type GMFCS Visual function GA at birth (wk) IQ level CBT MC NAI HSS-RPV HSS-LPV GS-HR GS-HL GS-Tot 1 6 F H PWM I 0 30 Norm 3 3.33 0.096 3 0 4.5 0 10.5 2 6.17 M H CDGM I 0 38 Norm 3.66 3 1.268 0 3 0 9 16 3 7.17 F H CDGM I 0 37 Norm 5.33 5 0.991 0 1.5 0 5.5 9.5 4 7.5 M H CDGM II 1 35 Border 4 3.33 0.950 0 2 0 9 17 5 8.33 F H BM I 0 40 Border 4.33 3.33 1.232 0 0 0 6 6 6 8.75 F H CDGM I 1 37 Norm 4.66 4.66 0.937 0 1 0 8 13 7 8.92 M H BM I 0 39 Norm 4.33 4 0.003 0 0 0 6 6 8 9.33 M H PWM II 0 40 Mild 6 4.66 0.208 1 0 2.5 0 8.5 9 12.08 M H PWM I 0 41 Norm 7.66 8 4.529 0 4 0 7 12 10 12.17 M H BM I 0 41 Border 5.66 4 1.134 0 0 5.5 0 5.5 11 5.17 M D PWM II 2 30 Mild 2.33 3 0.131 3.5 3.5 5.5 5.5 12 12 7 M D PWM I 0 34 Norm 3.66 4.33 1.183 3.5 4 5.5 6.5 15 13 7.67 M D PWM I 1 30 Border 3.66 3.33 0.660 3.5 3.5 6 6 15 14 9.33 F D PWM II 1 33 Mild 3.33 3 0.986 4 4 7 7.5 18.5 15 11.42 F D PWM II 2 40 Mild 4.33 4 0.003 1.5 0 4 0 6 16 11.5 F D PWM II 1 28 Mild 3 4.33 1.747 2.5 2.5 3.5 3.5 9 17 12 F D PWM II 2 36 Mild 4 4 0.285 4 4 6.5 7 19.5 GMFCS, Gross Motor Function Classification System; GA, gestational age; CBT, Corsi Block-tapping Task; MC, Magic Carpet span; NAI, navigational advantage index; HSS, hemispheric subscore; RPV, right periventricular; LPV, left periventricular; GS, global score; HR, hemispheric right; HL, hemispheric left; Tot, total; PWM, periventricular white matter; Norm, normal; CDGM, cortical and deep grey matter; Border, borderline; BM, brain maldevelopments; Mild, mild intellectual disability. 32 Developmental Medicine & Child Neurology 2015, 57 (Suppl. 2): 31 36

(a) (b) Figure 1: (a) Example images from the magnetic resonance imaging scans (fluid-attenuated inversion recovery [FLAIR] sequences) of patient 12. Note the typical lesion involving periventricular and middle white matter on both hemispheres, extending to all four major lobes. The corpus callosum, visible in the sagittal scan at the bottom left, is of reduced thickness in its middle and posterior parts, whereas the cerebellum is spared. (b) Classification form relative to the images shown, according to Fiori et al. 12 The lesion, which is apparent in four out of six axial slices (encircled labels b d and e) and in the corpus callosum, is outlined in dark grey. See text and Fiori et al. 12 for further details about the scoring procedure. graphical template divided into six axial brain slices, plus others for basal ganglia, brainstem, corpus callosum, and cerebellum. This enabled quantification of the involvement of each of three major layers (periventricular, middle, and cortico-subcortical) within each lobe (frontal, parietal, temporal, and occipital). A raw score was given to each layer on each lobe, as the number of slices where it was involved divided by the number of slices where it was represented. Raw scores were then transformed into semi-quantitative subscores: 0 (no involvement), 0.5 (up to half of the slices involved), or 1 (more than half of the slices involved). The sum of layer subscores on each lobe gave the lobar subscore, ranging from 0 to 3. The hemispheric global score (0 12) was the sum of lobar subscores on that hemisphere. The total global score (0 40) was the sum of right and left hemispheric scores, plus the global scores for basal ganglia and brainstem (0 10), corpus callosum (0 3), and cerebellum (0 3). The scale has demonstrated high interrater reliability (intraclass correlation coefficient 0.92) and intrarater reliability (intraclass correlation coefficient 0.91). 12 The relationships between MRI scores and functional scores were studied by Spearman s correlation. The study was approved by the local Ethics Committee of the IRCCS Fondazione Stella Maris, Pisa (protocol number 07/2012). RESULTS Raw CBT and Magic Carpet span scores are reported in Table I, with the main scores of MRI classification and patient characteristics, and are shown in Figure 2. CBT span in children with CP (mean=4.29, SD=1.29) was significantly lower than in matched (mean=5.08, SD=1.25, t[16]= 2.7088, p=0.015) and non-matched comparisons (mean=5.34, SD=0.92, t[19.14]= 3.19, p=0.005). Magic Carpet span, on the contrary, did not significantly differ between children with CP (mean=4.08, SD=1.19) and both matched (mean=4.41, SD=0.85, t[16]= 1.08, p=0.295) and non-matched comparisons (mean=4.47, SD=0.69, t[18.04]= 1.30, p=0.208). Nor did NAI statistically differ between children with CP (mean=0.18, SD=1.44) and both matched (mean=0.06, SD=0.98, t[16] =0.25, p=0.802), and non-matched comparisons (mean= 0.10, SD=0.94, t[18.62]=0.75, p=0.463). CBT span was lower in bilateral spastic CP (mean=3.47, SD=0.66) than in unilateral CP (mean=4.86, SD=1.34, t[13.84]= 2.82, p=0.014) and in matched comparisons (t[6] = 2.78, p=0.032), whereas it did not differ between unilateral CP and matched comparisons (t[9]= 1.25, p=0.242). Magic Carpet span did not significantly differ across groups, neither between bilateral (mean=3.71, SD=0.59) and unilateral CP (mean=4.33, SD=1.46), nor between Cognitive Strategies for Locomotor Navigation Vittorio Belmonti et al. 33

(a) CBT span (number of blocks) (c) CBT carpet span (age-normalized) 4 3 2 1 0 1 2 3 4 5 6 7 8 5 6 7 8 9 10 11 12 Age (years) (b) Magic carpet span (number of tiles) (d) Magic carpet span (age-normalized) 2 0 2 4 2 3 4 5 6 7 8 5 6 7 8 9 10 11 12 Age (years) 0 1 2 3 4 0 HSS-RPV 1 2 3 4 5 6 7 GS-HR Figure 2: (a, b) Raw span scores obtained on the Corsi Block-tapping Task (CBT) (a) and on the Magic Carpet (b) by typically developing children (grey dots) and children with CP (filled squares: bilateral CP, all with periventricular white matter anomalies; half-filled squares: unilateral CP with periventricular white matter anomalies, lesion side is contralateral to the coloured half; half-filled circles: unilateral CP with cortical and deep grey matter anomalies, side indicated as above; half-filled triangles: unilateral CP with BM anomalies, side as above). Regression lines and 95% prediction intervals fitted on control data are shown in grey. (c, d) Main correlations found for the age-normalized CBT (c) and Magic Carpet (d) span scores. Right periventricular hemispheric subscore (HSS-RPV) correlates with CBT and right hemispheric global score (GS-HR). bilateral CP and matched comparisons (t[6]= 1.56, p=0.170), or between unilateral CP and matched comparisons (t[9]=0.00, p=0.998). NAI did not differ across CP forms and comparisons. No significant effect of lesion type was found. Only NAI seemed to differ, although not significantly, across lesion types, being higher in periventricular white matter lesions (mean=0.57, SD=1.61) than in cortical and deep grey matter lesions (mean= 0.07, SD=1.20) and in brain maldevelopments (mean= 0.79, SD=0.69). Normalized CBT span positively correlated with gestational age at birth (q=0.62, p=0.009). Moreover, it was significantly related to visual function impairment, as revealed by ANOVA (F 2,14 =5.15, p=0.021). A Tukey post-hoc test revealed a significant difference between levels 0 and 2 (p=0.044), and almost a significant difference between levels 0 and 1 (p=0.065). Neither normalized Magic Carpet span nor NAI correlated with either gestational age or visual functions. Right hemispheric global score significantly and inversely correlated both with CBT normalized span (q= 0.68, p=0.002) and with Magic Carpet normalized span (q= 0.54, p=0.025) (Fig. 2). NAI did not correlate with right hemispheric global score (q= 0.09, p=0.734). Neither left hemispheric global score nor total global score significantly correlated with any functional score. 34 Developmental Medicine & Child Neurology 2015, 57 (Suppl. 2): 31 36

A strong correlation was found between the hemispheric subscore for right periventricular white matter and normalized CBT span (q= 0.77, p<0.001) (Fig. 2), whereas it did not correlate with normalized Magic Carpet span (q= 0.34, p=0.185) or with NAI (q=0.10, p=0.693). The hemispheric subscore for right periventricular white matter was also negatively related to gestational age at birth (q= 0.65, p=0.005). None of the other MRI scores significantly correlated with any functional score; there was only an almost significant negative correlation between NAI and the sum of left and right temporal cortico-subcortical and middle white matter scores (q= 0.46, p=0.061). DISCUSSION This is the first study, to our knowledge, to explore a real navigation task in CP. In fact, despite a previous report of visual navigation deficits in adolescents with periventricular leukomalacia, 13 no cognitive task requiring real body motion has been used. The Magic Carpet has thus proved an innovative tool for assessing navigation in children with motor disorders. Our main finding is that, across children with forms of spastic CP, the distribution of impairment of spatial memory in reaching space (CBT) is different from that in navigational space (Magic Carpet). In general, CBT span is increasingly frequently reduced compared with Magic Carpet span. Although CBT span proved significantly lower than in the comparison group, this was not the case for Magic Carpet span. Only three children had an agenormalized Magic Carpet span under 2 SE, and none scored under 3 SE, whereas five had a normalized CBT span under 2 SE. Therefore, spatial memory in children with spastic CP seems to be more prominently impaired in reaching than in navigational space. This relative sparing of navigation performance is somewhat in contradiction with the notion that navigation requires more advanced and complex cognitive skills than basic visual spatial memory. However, more complexity also means more factors involved and more possible mechanisms for compensation, both at the neural and cognitive levels. In fact, several strategies and brain networks can be used to solve a task like the Magic Carpet. As previously shown, 10 the Magic Carpet fosters a switch from ego- to allocentric spatial encoding, but it can also be solved, although less efficiently, by egocentric updating. These strategies rely on different brain networks, the former centred on hippocampal and prefrontal areas, the latter on posterior parietal and premotor cortices. 4 Alternatively, or in addition, visual spatial impairment in children with CP could pertain specifically to egocentric reference frames, as it typically occurs in hemispatial neglect. 4 This hypothesis would be coherent with dorsal-stream vulnerabilty, because egocentric encoding is mostly associated with dorsalstream activation. 14 Finally, it could be just a matter of scale: larger stimuli are better perceived and stored by children with visual defects. The strongest (negative) correlation we have found is that between CBT and right periventricular white matter injury. CBT span was also related to global right hemispheric impairment, which, on the other hand, was the only determinant of Magic Carpet span. We therefore confirm the tight relationship between visual spatial impairment and deep white matter damage, and the more general association between spatial functions and the right hemisphere. In fact, although both right- and left-dominated strategies have been demonstrated for navigation, 15 the latter are based on route memory (encoding a sequence of turns), which is not available on the Magic Carpet, as sequences are visually delivered. Right-only correlations might also depend on left lesions being more heterogeneous (all four cortical and deep grey matter lesions and two out of three brain maldevelopment lesions were on the left side). In addition, CBT span proved positively related to gestational age at birth, but, of course, a low gestational age is also a determinant of white matter impairment. The relative contributions of periventricular lesions and of preterm birth to spatial disorders are usually hard to disentangle (see Ego et al., this issue). Visual function impairment was another, partly independent, determinant of CBT. Importantly, no direct relationship with visual function was found for navigation. As mentioned above, the latter point can have at least two explanations. At a low, perceptual level, the Magic Carpet delivers larger, and thus better perceived, stimuli. At a higher, cognitive level, navigational space activates a broader, multimodal network than reaching space, thus resulting in less sensitivity to visual defects. 4 NAI differences seemed to indicate a trend towards relatively better navigational skills from brain maldevelopments to cortical and deep grey matter lesions, to periventricular white matter lesions. This trend, however, is more attributable to CBT span reduction due to white matter injury and low gestational age than to differences in Magic Carpet span. The same consideration can be made for the correlation between NAI and temporal cortico-subcortical impairment: although the role of medial temporal lobe structures in navigation is well known, this finding seems more attributable to CBT span reduction in periventricular white matter lesions. To conclude, navigation is an almost unexplored topic in the field of CP and early brain lesions. It is not an abstract cognitive activity, but is involved in many common and vital tasks such as generating efficient walking trajectories, an apparently simple but actually demanding and lifelong developing ability. 16 Being more complex and less visually dependent than basic visual spatial memory, navigation could be more suitable for compensation and remediation strategies. Finally, promoting the early exploration of navigational space could enable a better, multisensory representation of spatial relationships and shapes, which could be then generalized across settings and spaces. Cognitive Strategies for Locomotor Navigation Vittorio Belmonti et al. 35

ACKNOWLEDGEMENTS This research was supported by a PACE grant from La Fondation Motrice Sodiaal (Paris, France) to AB and GC, by a MED/39 grant from Regione Toscana (financed by the Post-doctoral Fellowships Programme ESF 2012) and University of Pisa to VB, and by scholarship support to VB from La Fondation Motrice. The funders did not take part in study design, data collection, data analysis, manuscript preparation, or publication decisions. The authors have stated that they had no interests which might be perceived as posing a conflict or bias. REFERENCES 1. Atkinson J, Braddick O. From genes to brain development to phenotypic behavior: dorsal-stream vulnerability in relation to spatial cognition, attention, and planning of actions in Williams syndrome (WS) and other developmental disorders. Prog Brain Res 2011; 189: 261 83. 2. Nemmi F, Boccia M, Piccardi L, Galati G, Guariglia C. Segregation of neural circuits involved in spatial learning in reaching and navigational space. Neuropsychologia 2013; 51: 1561 70. 3. Burgess N. Spatial memory: how egocentric and allocentric combine. Trends Cogn Sci 2006; 10: 551 7. 4. Galati G, Pelle G, Berthoz A, Committeri G. Multiple reference frames used by the human brain for spatial perception and memory. Exp Brain Res 2010; 206: 109 20. 5. Newcombe NS, Ratliff KR, Shallcross WL, Twyman AD. Young children s use of features to reorient is more than just associative: further evidence against a modular view of spatial processing. Dev Sci 2010; 13: 213 20. 6. Corsi PM. Human memory and the medial temporal region of the brain (Ph.D. Thesis). Montreal: McGill University, 1972: 84. 7. Pagulayan KF, Busch RM, Medina KL, Bartok JA, Krikorian R. Developmental normative data for the Corsi Block-tapping Task. J Clin Exp Neuropsychol 2006; 28: 1043 52. 8. Gagliardi C, Tavano A, Turconi AC, Pozzoli U, Borgatti R. Sequence learning in cerebral palsy. Pediatr Neurol 2011; 44: 207 13. 9. Piccardi L, Iaria G, Ricci M, Bianchini F, Zompanti L, Guariglia C. Walking in the Corsi test: which type of memory do you need? Neurosci Lett 2008; 432: 127 31. 10. Belmonti V, Cioni G, Berthoz A. Switching from reaching to navigation: differential cognitive strategies for spatial memory in children and adults. Dev Sci 2014; Published online 28 November 2014, doi: 10.1111/ desc.12240 (E-pub ahead of print). 11. Kr ageloh-mann I, Horber V. The role of magnetic resonance imaging in elucidating the pathogenesis of cerebral palsy: a systematic review. Dev Med Child Neurol 2007; 49: 144 51. 12. Fiori S, Cioni G, Klingels K, et al. Reliability of a novel semi-quantitative scale for classification of structural brain magnetic resonance imaging in children with cerebral palsy. Dev Med Child Neurol 2014; 56: 839 45. 13. Pavlova M, Sokolov A, Kr ageloh-mann I. Visual navigation in adolescents with early periventricular lesions: knowing where, but not getting there. Cereb Cortex 2007; 17: 363 9. 14. Committeri G, Galati G, Paradis A-L, Pizzamiglio L, Berthoz A, Le Bihan D. Reference frames for spatial cognition: different brain areas are involved in viewer-, object-, and landmark-centered judgments about object location. J Cogn Neurosci 2004; 16: 1517 35. 15. Igloi K, Doeller CF, Berthoz A, Rondi-reig L, Burgess N. Lateralized human hippocampal activity predicts navigation based on sequence or place memory. Proc Natl Acad Sci U S A 2010; 107: 14466 71. 16. Belmonti V, Cioni G, Berthoz A. Development of anticipatory orienting strategies and trajectory formation in goal-oriented locomotion. Exp Brain Res 2013; 227: 131 47. 36 Developmental Medicine & Child Neurology 2015, 57 (Suppl. 2): 31 36