Loss of normal facial asymmetry in schizophrenia and bipolar disorder: Implications for development of brain asymmetry in psychotic illness

Development of the craniofacies occurs in embryological intimacy with development of the brain and both show normal left-right asymmetries. While facial dysmorphology occurs to excess in psychotic illness


Introduction
Within a lifetime trajectory model, concepts of an early neurodevelopmental component in risk for psychotic illness continue to evolve at multiple levels of enquiry (Murray and Lewis, 1987;Weinberger, 1987Weinberger, , 2017;;Waddington, 1993;McGrath et al., 2003;Waddington et al., 2012;Murray et al., 2017;Howes and Shatalina, 2022).Biological evidence for the importance of developmental disruption during fetal life comes from an excess in schizophrenia of major congenital anomalies and dysfunctions (Waddington et al., 2008) and particularly minor physical anomalies, among which facial dysmorphologies are the most common (Lane et al., 1997;Xu et al., 2011;Bora, 2022) and share the closest embryological relationship with fetal brain development (Schneider et al., 2001;Marcucio et al., 2015;Naqvi et al., 2021).Thus, facial anomalies are long recognised to be a putative index of disruption to processes regulating early brain development (DeMyer et al., 1964;Kjaer, 1995;Waddington et al., 1999).
However, as facial development is an intrinsically 3-dimensional (3D) processes, craniofacial dysmorphologies in psychotic illness cannot be fully informative on underlying pathobiological processes unless analysed in 3D.On application of 3D laser surface imaging and geometric morphometrics to analyse craniofacial dysmorphology in schizophrenia we have resolved the topography of overall dysmorphology to those facial regions, the frontonasal prominences (Hennessy et al., 2007(Hennessy et al., , 2010)), that show the greatest embryological intimacy with brain morphogenesis (Schneider et al., 2001;Marcucio et al., 2015;Naqvi et al., 2021).
In the context of evolving spectrum-dimensional concepts of psychotic illness (Barch et al., 2013;Guloksuz and van Os, 2018;Altinbas et al., 2021;Nkire et al., 2021), there is increasing pathobiological and genetic evidence for a close relationship between schizophrenia and bipolar disorder (Birur et al., 2017;Bora et al., 2018;Bipolar Disorder and Schizophrenia Working Group of the Psychiatric Genomics Consortium, 2018).On a background of continuing debate on the neurodevelopmental relationship between these two disorders (Parellada et al., 2017;Kloiber et al., 2020), minor physical anomalies are present in bipolar disorder to an extent approaching that evident in schizophrenia (Bora, 2022).We have also applied 3D laser surface imaging and geometric morphometrics to identify holistic craniofacial dysmorphology in bipolar disorder and find this to show commonalities with that evident in schizophrenia (Hennessy et al., 2010;Katina et al., 2020).
However, the studies outlined above fail to consider a distinct, more fundamental and potentially more incisive domain of vertebrate development: the extent to which morphogenesis proceeds symmetrically or involves the embryonic breaking of left-right symmetry to create asymmetry whereby there emerge quantitative differences between the left and right sides of a given structure (Grimes and Burdine, 2017;Little and Norris, 2021).For example, a cardinal feature of healthy subjects is brain asymmetry that includes the frontal lobes: the right is wider and protrudes anteriorly relative to the left and shows a slight leftward torque (Toga and Thompson, 2003;Zhao et al., 2022).Facial asymmetry is also evident in healthy subjects (Srivastava et al., 2018;Reddy et al., 2023) and we have developed new geometric morphometric techniques for its 3D analysis (Sukno et al., 2015).Here we apply these techniques, to our knowledge for the first time, to: (i) quantify and visualise the topography of facial asymmetry in healthy subjects; (ii) investigate whether such normal facial asymmetry is altered in schizophrenia and in bipolar disorder; and (iii) consider the embryological implications of any alterations in normal facial asymmetry for the still controversial issue of whether cerebral asymmetry is disrupted in schizophrenia and the extent to which this might apply also to bipolar disorder (DeLisi et al., 1997;Sommer et al., 2001;Crow et al., 2013;Sha et al., 2021a).

Participants
Approval for this study was obtained from the Research Ethics Committee of St. John of God Hospital, Stillorgan, Co. Dublin.All subjects were adults between the ages of 18-65 who gave written, informed consent to participation after the procedures had been fully explained.
The subjects of this study have been outlined in a previous report on holistic facial dysmorphology in schizophrenia and bipolar disorder (Katina et al., 2020) that did not address the independent domain of facial asymmetry that is the topic of the present report.Subjects were recruited from and representative of patients having a clinical diagnosis of schizophrenia or bipolar disorder who were admitted to St. John of God Hospital or treated as outpatients at its associated community mental health service and were experiencing either a first psychotic or manic episode, or a recent psychotic or manic relapse.Subjects were included where these diagnoses were confirmed using the Structured Clinical Interview for DSM-IV (First et al., 1995).Control subjects were recruited from the administrative, ancillary and clinical staff of St. John of God Hospital and its associated community mental health service who, on interview, reported no personal or family history of psychosis, major affective disorder or suicide.To exclude ethnic differences in craniofacies, all subjects reported on interview that their parents and grandparents were born on the island of Ireland [Republic of Ireland or Northern Ireland], England, Wales or Scotland; all subjects were white.Additionally, subjects were excluded if, on interview, they reported craniofacial trauma or surgery, or had a beard or moustache.There were 49 control subjects [mean age 33.2 (standard deviation (SD) 10.9); 20 males, 29 females], 22 subjects with schizophrenia [mean age 33.1 (12.3); 18 males, 4 females] and 22 subjects with bipolar disorder [mean age 35.9 (10.6); 13 males, 9 females].

3D laser surface imaging and image processing
As described previously in detail (Hennessy et al., 2007(Hennessy et al., , 2010;;Prasad et al., 2015;Katina et al., 2020), facial surfaces were recorded by a single investigator (BK) using a portable, hand-held 3D laser imaging system (Polhemus FastScan, Vermont, USA); typical surfaces (Fig. 1) consisted of ≈80,000 points.Incomplete data due to hair and complex folded surfaces were processed using a fully automated algorithm (Hu et al., 2012) that we have shown to achieve an optimal balance between performance and triangle manipulation (Rojas et al., 2014).

Geometric morphometric and statistical analysis of facial asymmetry
We obtained point-wise estimates of facial asymmetry (i.e.differences in position between the left and right hemispheres for each point on the facial surface; Fig. 1) directly in physical units by alignment of the facial surface with a mirrored-reflected version of itself.To this end, we used a least median of squares (LMedS) algorithm with weights that decay exponentially with distance from the midline (distance-weightedhemispheres and midline-LMedS (DW-HM-LMedS); this DW-HM-LMedS approach has been shown to perform considerably better than competing alternatives in terms of estimation accuracy and, more importantly, to produce estimations of facial asymmetry patterns that are highly correlated with actual asymmetry (for details see Sukno et al., 2015; a technical summary is included as Supplementary material).To allow statistical comparisons, principal component analysis (PCA; Hennessy et al., 2007Hennessy et al., , 2010;;Prasad et al., 2015;Katina et al., 2020) was F.M. Sukno et al. Psychiatry Research 342 (2024) 116213 applied, followed by parallel analysis (Franklin et al., 1995) that, at the specified overall level of significance (p < 0.05), identified a set of nine components of asymmetry that explained 87 % of variation in asymmetry.
As these extracted components represent only asymmetry, they are conceptually distinct from the holistic (symmetrized) dysmorphology described previously (Hennessy et al., 2007(Hennessy et al., , 2010;;Prasad et al., 2015;Katina et al., 2020).Therefore, asymmetries are independent of, and can be quantified and visualised separately from, holistic aspects of facial shape, which can then be chosen to be the same for each group to facilitate biological interpretation.In this work we use the symmetrized average facial shape as the basis on which to display asymmetry patterns.
Statistical analysis involved one multivariate analysis of covariance (MANCOVA) across the nine principal components (PCs) of asymmetry for the three diagnostic groups, with diagnosis and sex as factors and age as a covariate; the model was therefore prediction of PC-scores by diagnosis, sex, age and the interaction of diagnosis and sex.Comparisons between the three diagnostic groups were performed using Hotelling's T 2 test (Hennessy et al., 2007(Hennessy et al., , 2010; a technical summary is included as Supplementary material) with statistical significance indicated by Bonferrini-corrected p < 0.05 [i.e.absolute p < 0.05/3 = 0.0166].Analyses were performed using SPSS version 23.

Visualisation of overall asymmetries
These statistically significant differences across and between control, schizophrenia and bipolar disorder groups were given biological import first through visualisations of individual group asymmetries.In each figure the right (R) side of the subject is shown on the left (L) side of the image, in accordance with anatomical-radiological convention.
Each of the nine significant components identified by PCA can be interpreted as patterns of asymmetry over the facial surface that indicate the extent to which any given point differs from a perfectly symmetric face.Thus, the extent of asymmetry of a given point on the left side of the face will be the same as the extent of asymmetry of its counterpart on the right side, but in the opposite direction.This is also true for the combination of these patterns, whether as a whole (i.e. the nine components together within each subject group) or for differences between these groups.This allows for the display of meaningful visualisations of both the overall asymmetries per group (i.e.control, schizophrenia and bipolar disorder) and the asymmetry differences between these groups.
Overall asymmetries as analysed in MANCOVA across the nine PCs of asymmetry resolved for each of the three groups are shown in Fig. 2. For completeness, data are shown initially as Euclidean distances from pure symmetry at each point on the facial surface.Magnitude of asymmetry is colour-coded according to the scale (distance in mm from symmetry) signed with respect to the surface normal, i.e. the perpendicular vector to the local surface area: positive values (red scale) indicate distances outward of the surface normal and negative values (blue scale) indicate distances inward of the surface normal.We focus here on bilateral features of the upper face/periorbital regions, which enjoy the greatest embryological intimacy with and spatial proximity to the brain.In the analysis of asymmetry: (i) their lateral (outer) surfaces on the right (Fig. 2, 1R) are compared with their counterpart surfaces on the left (1 L); (ii) their dorsal (superior)-medial (inner) surfaces on the right (2R) are compared with their counterpart surfaces on the left (2 L); (iii) their ventral (inferior)-medial (inner) surfaces on the right (3R) are compared with their counterpart surfaces on the left (3 L).
Using regional terminology for phenotypic variations in craniofacial topographies from Elements of Morphology (Allanson et al., 2009;Prasad et al., 2015;Katina et al., 2020), these analyses and visualisations  (Fig. 2) reveal asymmetries to be most prominent for the upper face and periorbital regions that enjoy the greatest embryological intimacy with and spatial proximity to the brain; these are described in detail below.They are also prominent for the mandible and chin, which are outlined further below.Asymmetries in mid-facial regions (premaxilla and midface; nose and filtrum; lips, mouth and oral regions) are subtle, hence they are included in Figs. 2 and 3 but are not described in detail.

Upper facial and periorbital region
On visualising the findings on Hotelling's T 2 tests following MAN-COVA (see Section 3.1.),in controls there was: (i) prominent displacement outward from the surface of the right lateral forehead and periorbital region (Fig. 1, area 1R in red) relative to its corresponding left region (1 L in blue); (ii) prominent outward displacement from the left dorsomedial forehead and periorbital region (2 L in red) relative to its corresponding right region (2R); (iii) localised outward displacement from the right ventromedial forehead and periorbital region (3R in orange) relative to its corresponding left region (3 L in turquoise).In schizophrenia there was substantial diminution in these asymmetries.In bipolar disorder there was diminution in, but residual retention of, this control pattern of asymmetries.

Visualisation of inter-group differences in asymmetry along orthogonal axes
The above analyses and their visualisation in Fig. 2 present distances from pure symmetry for each point on the facial surface.Thus, they delineate comprehensively the overall craniofacial (a)symmetries for each of the three individual groups.While the directional diversity of the visualisations in Fig. 2 accurately reflects the real-world biology of each group, abrupt transitions from (for example) red to blue, do not necessarily indicate abrupt reversal of spatial relationships; this is because a small shift in the angle between the direction of asymmetry and the surface normal at a given location from 90 • to 89 • (i.e.− 1 • ) or to 91 • (i.e. +1 • ) could result in such abrupt reversal of colour coding.Specifically, this mode of display does not specify differences in asymmetries between groups along orthogonal x-, y-, z-axes that, by convention, define anatomy.
Therefore, the visualisations in Fig. 3a present decomposition of the overall asymmetries evident in the control group (as displayed in Fig. 2).These display Euclidean distances from pure symmetry along x-, y-and z-axes (Fig. 1) for each point on the facial surface, colour-coded in mm: positive values (red scale) indicate asymmetries as displacements in the directions x [R-L], y [ventral (V)-dorsal (D)] and z [posterior (P)anterior (A)] in controls; negative values (blue scale) indicate displacements in the opposite directions.The visualisations in Fig. 3b/c then present decompositions of those overall patterns of asymmetry (as displayed in Fig. 2) that differ significantly between groups [schizophrenia vs controls, p < 0.001; bipolar disorder vs controls, p = 0.027; see 3.1.].As the absolute magnitudes of mean differences in asymmetries between groups are relatively small, their corresponding visualisations are subtle.Therefore, in order to present their import more effectively, the displayed images for inter-group differences magnify the asymmetries for each group by a factor of 5.

x-axis
In controls the dorsal-medial area (Fig. 3a, blue area 4) is asymmetrically displaced to the right; below this is a ventral-lateral area (Fig. 3a, orange area 5R/L) that is asymmetrically displaced to the left on both sides.In schizophrenia (Fig. 3b) the dorsal-medial area 4 (now red) is displaced to the left relative to controls, indicating reduction in the rightward asymmetry evident in controls; the ventral-lateral area 5 (now turquoise) is displaced to the right relative to controls, indicating reduction in the leftward asymmetry evident in controls.In bipolar disorder (Fig. 3c) the dorsal-medial area 4 (now orange) is displaced to the left relative to controls, indicating reduction in the rightward asymmetry evident in controls; the ventral-lateral area 5 (now turquoise) is displaced to the right relative to controls, indicating reduction in the leftward asymmetry evident in controls.However, the magnitude of these reductions in extent of control asymmetry in bipolar disorder is less than that evident in schizophrenia.

y-axis
In controls the right lateral area (Fig. 3a, blue area 6R) is asymmetrically displaced ventrally relative to its left counterpart (red area 6 L).In schizophrenia (Fig. 3b) the right lateral area 6R (now red) is displaced dorsally relative to its left counterpart 6 L (now blue), indicating reduction in the control pattern of asymmetry.In bipolar disorder (Fig. 3c) the right lateral area 6R (now orange) is displaced dorsally relative to its left counterpart 6R (now turquoise), indicating reduction in the control pattern of asymmetry.However, the magnitude of these reductions in extent of control asymmetry in bipolar disorder is less than that evident in schizophrenia.

z-axis
In controls the right outer lateral area (Fig. 3a, red area 7R) is asymmetrically displaced anteriorly relative to its left counterpart (blue area 7 L).In schizophrenia (Fig. 3b) the right outer lateral area 7R (now blue) is displaced posteriorly relative to its left counterpart 7 L (now red), indicating reduction in the control pattern of asymmetry.In bipolar disorder (Fig. 3c) the right outer lateral area 7R (now blue) is displaced posteriorly relative to its left counterpart 7 L (now red), indicating reduction in the control pattern of asymmetry.The magnitude of these reductions in extent of control asymmetry in bipolar disorder is similar to that evident in schizophrenia.Additionally, in schizophrenia the right medial area (Fig. 3b, red area 8R) is asymmetrically displaced anteriorly relative to its left counterpart (blue area 8 L), with no such asymmetry being evident in controls or in bipolar disorder (Fig. 3a/c).

Mandible and chin
In controls there is prominent asymmetry of the chin characterised by displacements inward from the surface on the right (blue) relative to its left counterpart (red; Fig. 2); this direction of asymmetry is opposite to that of the lateral upper face and periorbital region (areas 1R/L) and similar to that of its dorsal-medial region (areas 2R/L).In schizophrenia and particularly in bipolar disorder the most ventral aspect of the chin shows reversal of this control pattern of asymmetry (Fig. 2).Decomposition of these asymmetries in controls along orthogonal axes (Fig. 3a) indicates that along the x-axis the ventral-medial chin (blue) is displaced to the right and along the z-axis the right lateral chin (blue) is displaced posteriorly.Regarding statistically significant differences in asymmetry patterns in patient groups relative to controls along orthogonal axes (Fig. 3b/c), in schizophrenia along the x-axis the ventral-medial chin (dark blue) is displaced further to the right and along the z-axis the right ventral chin (red) is displaced anteriorly; in bipolar disorder along the xaxis the lateral chin (blue) is displaced to the right on both sides and along the z-axis the right lateral chin (red) is displaced anteriorly.

Main findings
In this first study of facial asymmetry in psychotic illness, 3D laser surface imaging and geometric morphometric analysis indicates (a) In control subjects the upper face and periorbital region show marked overall asymmetries whose geometry includes: along the x-axis, rightward asymmetry in its dorsal-medial aspects and leftward asymmetry in its ventral-lateral aspect; along the y-axis, ventral asymmetry in its right dorsal-lateral aspect; along the z-axis, anterior protrusion in its right ventral-lateral aspect.(b) In both schizophrenia and bipolar disorder these normal facial asymmetries are diminished, this diminution being most evident in schizophrenia with some residual retention of asymmetries in bipolar disorder.

Embryology of upper face-brain development in relation to psychosis
The face and brain share a common embryological origin in neural crest cells that during early fetal life migrate from ectodermally derived primordia to regulate morphogenesis via both molecular signalling and architecture (Schneider et al., 2001;Marcucio et al., 2015;Naqvi et al., 2021); thus, facial anomalies are long recognised to accompany disruption to processes regulating early brain development (DeMyer et al., 1964;Kjaer, 1995;Waddington et al., 1999).This co-development is most intimate in relation to the frontonasal ectodermal zone that directs what will become the frontonasal facial prominences and frontal brain regions (Schneider et al., 2001;Marcucio et al., 2015;Naqvi et al., 2021).
We have reported a characteristic topography of frontonasal dysmorphology in schizophrenia (Hennessy et al., 2007(Hennessy et al., , 2010)), in bipolar disorder (Hennessy et al., 2010;Katina et al., 2020) and in 22q11.2deletion syndrome (Prasad et al., 2015) which is associated with markedly increased risk for psychosis.On the basis of the embryological timeline of face-brain relationships over fetal life (Diewert and Lozanoff, 1993;Diewert et al., 1993), which do not consider issues of asymmetry, we proposed that the pathobiology of these psychotic disorders includes dysmorphogenic event(s) during gestational weeks (GW) 6 through 19-20 (Hennessy et al., 2007;Katina et al., 2020).However, these studies were holistic and did not evaluate the fundamentally distinct domain of morphogenesis that involves embryonic breaking of left-right symmetry to create asymmetry, i.e. the development of quantitative differences between the left and right sides of a given structure.

Upper facial-periorbital asymmetries in relation to brain asymmetries
The primary geometry of normal facial asymmetries identified here, in areas of the upper facial and periorbital regions that enjoy exquisite embryological and spatial intimacy with the forebrain, shows notable commonalities with the primary geometry of normal frontal lobe asymmetries: along the x-axis the right frontal lobe is wider relative to the left; along the z-axis the right frontal lobe protrudes anteriorly relative to the left (Toga and Thompson, 2003;Zhao et al., 2022); along the y-axis frontal lobe asymmetries are more subtle (Zhao et al., 2022), with facial development attaining greater ventral separation from the brain relative to the x-and z-axes (Schneider et al., 2001;Marcucio et al., 2015).Therefore, the loss of these normal facial asymmetries in subjects with schizophrenia and bipolar disorder may inform on controversies regarding disruption of cerebral asymmetry in these disorders (DeLisi et al., 1997;Sommer et al., 2001;Crow et al., 2013;Sha et al., 2021a).

Embryology of facial asymmetry in relation to psychosis
The present analyses involve resolution of the geometry of facial asymmetry, independent of holistic (i.e.symmetrised) dysmorphology.Though the embryonic face-forebrain does not evidence consistent asymmetric gene expression at GW5-5.5, such asymmetries become evident across GW7.5-13, during which the right anterior cortex shows higher expression of genes that also increase across early fetal life, i.e. is the 'faster' side (de Kovel et al., 2018;Sha et al., 2021b).Thus, the present disruption to asymmetries of the upper face and periorbital region in both schizophrenia and bipolar disorder, and by embryological inference to asymmetries of the frontal lobe, implicates dysmorphogenic event(s) that include GW7-14.

Embryology of mandibular-chin asymmetries in relation to psychosis
Mandibular-chin regions are considerably less related embryologically to the brain than are the frontonasal regions (Schneider et al., 2001;Marcucio et al., 2015;Abramyan and Richman, 2018), with genetic face-brain correlations in healthy subjects strongest for the upper face/periorbital regions/frontonasal prominences and weakest for premaxilla and mandible (Naqvi et al., 2021).Neural crest cells also migrate out from those related to the hindbrain and regulate morphogenesis of the mandibular region (Schneider et al., 2001;Marcucio et al., 2015).Mutually exclusive frontal vs hindbrain gene-expression patterns are established by GW14-17, with cells between these cortical poles less specified towards regional identity (Bhaduri et al., 2021).The present findings that normal facial asymmetries and their disruption in schizophrenia and bipolar disorder involve primarily upper facial-periorbital and mandibular-chin regions, and less so for regions between these facial poles, may therefore relate to such processes and their disruption in psychotic illness.

Strengths and limitations
In this study we applied novel procedures to analyse facial asymmetry in 3D and to systematically compare schizophrenia and bipolar subjects with controls.Artefacts associated with facial differences due to ethnicity and trauma were minimised.Clinical imaging, image processing, geometric morphometric analysis and biological interpretation were performed by independent groups of investigators who were each blind to the others' findings until the study was completed.
There was incomplete imaging of the most dorsal aspects of the upper forehead due to hairline.While we did not obtain information on handedness, evidence (Toga and Thompson, 2003;Zhao et al., 2022) indicates that forebrain asymmetries may be slightly more evident in right-than in left-handed persons and in males than in females, with no handedness × sex interaction.The present effect of age on facial asymmetry is similar to that reported for brain asymmetry and interpreted in terms of the interplay of differential growth dynamics across left and right hemispheres (Toga and Thompson, 2003;Zhao et al., 2022).While antipsychotic-antimanic drugs may cause weight gain and thus slightly alter overall facial shape, there is no evidence for asymmetric effects.
The extent to which these findings may generalise to larger numbers of male and female subjects across ethnicities and may relate to aspects of psychopathology and cognitive dysfunction remains to be determined.While subjects with bipolar disorder were not subdivided into those 'with' vs 'without' psychotic features, we have recently reported that these putative subgroups reflect dichotomization at a subjective threshold along a continuous, normally distributed dimension of psychosis severity intrinsic to bipolar disorder (Nkire et al., 2024).As these studies did not include cranial neuroimaging to resolve both brain and facial structure (Hostalet et al., 2024), interpretation of facial vis-à-vis brain asymmetries remains indirect and is predicated on their embryological relationship.

Conclusions
The present findings indicate marked facial asymmetries in control subjects, with loss of asymmetries in the upper face-periorbital area in schizophrenia.Embryologically, they implicate disruption to facial morphogenesis and establishment of geometric asymmetry particularly during GW7-14 and are consistent with still controversial loss of brain asymmetry in schizophrenia (DeLisi et al., 1997;Sommer et al., 2001;Crow et al., 2013;Sha et al., 2021a).This loss of facial asymmetry was evident also in bipolar disorder, indicating commonality with schizophrenia in the underlying processes.

Funding statement
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Fig. 1 .
Fig. 1.Representative facial surface from 3D laser imaging.The x-, y-and zaxes applied to each point on the facial surface for geometric morphometric analyses and in Fig. 3 are indicated by superimposition on a coronal-sagittal oblique view.They correspond to left-right (x), dorsal-ventral (y) and anterior-posterior (z) directions relative to the anatomical axes of the face.

Fig. 2 .
Fig. 2. Visualisation of patterns of asymmetry for each of the three diagnostic groups: left panel, controls; centre panel, schizophrenia; right panel, bipolar disorder.Data are mean Euclidean distances from pure symmetry at each point on the facial surface, colour-coded in mm: positive values (red scale) indicate displacements outward of the surface; negative values (blue scale) indicate displacements inward of the surface.In the upper face and periorbital regions, the following surface areas are indicated: lateral [1 right/left (1R/L)], dorsal-medial [2R/L] and ventral-medial [3R/L].

Fig. 3 .
Fig. 3. Visualisation of patterns of asymmetry in controls along orthogonal axes and visualisation of significant differences in patterns of asymmetry between control and patient groups along orthogonal axes (see Fig. 1).(a) For controls data are mean Euclidean distances from pure symmetry along x-, y-and z-axes for each point on the facial surface, colour-coded in mm: positive values (red scale) indicate asymmetries as displacements in the directions x [right-left (R-L)], y [ventral-dorsal (V-D)] and z [posterior-anterior (P-A)]; negative values (blue scale) indicate displacements in the opposite directions.In the upper face and periorbital regions of controls the following surface areas are indicated: dorsal-medial [4] and ventral-lateral [5R/L] on x-axis, lateral [6R/L] on y axis, outer lateral [7R/L] on z-axis.(b) and (c) For patient groups data are shown as mean differences in Euclidean distances in asymmetry at each point on the facial surface for schizophrenia (SZ) relative to control (C) groups and bipolar disorder (BD) relative to control (C) groups along x-, y-and z-axes for each point on the facial surface, colour-coded in mm: positive values (red scale) indicate between-group differences in asymmetries as displacements in the directions x [R-L], y [V-D] and z [P-A]; negative values (blue scale) indicate displacements in the opposite directions.In the upper face and periorbital regions of schizophrenia subjects the following additional area is indicated: dorsalmedial [8R/L] on z-axis.