F172. INDIVIDUAL PREDICTION OF RISK IN ADOLESCENT OFFSPRING OF PARENTS WITH SCHIZOPHRENIA OR BIPOLAR DISORDER: A MACHINE LEARNING NEUROIMAGING STUDY WITH A CROSS-STAGE VALIDATION

Abstract Background Schizophrenia (SZ) and bipolar disorder (BD) are severe psychiatric disorders that are not easily distinguishable based on clinical measures. Offspring of patients with SZ or BD have a tenfold increased risk of developing the disorder as well as an increased risk for other severe mental disorders. Reliable identification of these subjects might allow for early recognition and intervention, which have been shown to be beneficial for treatment outcome and may even prevent transition to illness. Based on abundant evidence that SZ and BD are associated with structural brain abnormalities, we investigated whether MRI brain-scans can be used to detect individual risk of developing SZ or BD in adolescents. Methods Structural MRI brain-scans were acquired in adolescent offspring (8–19 year) of parents with schizophrenia (oSZ;N=50), bipolar disorder (oBD;N=82), and without a mood or psychotic DSM-IV disorder (oHC;N=53), as part of the Dutch Bipolar and Schizophrenia Offspring Study (DBSOS). Support vector machine (SVM) models were trained on the gray matter tissue density maps to predict to which offspring class (oHC/oBD/oSZ) an individual belonged. Prediction accuracy was assessed using cross-validation. To validate our prediction models, we applied them to the tissue maps from subjects from a sample of unrelated HC/BD/SZ adults. Secondly, validated prediction models built from the adult subjects’ MRI scans were applied to the tissue maps of the adolescents to predict illness class (HC/BD/SZ). Results The offspring-based model separated oHC/oSZ individuals with 77% accuracy (p<0.001), oHC/oBD with 68% accuracy (p<0.001), and oBD/oSZ with 64% accuracy (p<0.01). The adult-based models could separate the patients’ offspring from the healthy offspring with 66–70% accuracy, but oBD from oSZ with lower accuracy (59%). In addition, the offspring models could separate adult patients from control subjects with comparable accuracy (66–68%) and separate the two patient groups with moderate accuracy (69%). Discussion The familial high-risk adolescents could be separated from controls with moderate to high accuracy (up to 77%), based on their MRI-scans. Moreover, the brain tissue patterns based on risk (adolescents) or illness (adults) were able to predict (risk) class in the other stage group. These results show (1) that high-risk individuals already show brain abnormalities, and (2) display similarities with abnormalities in ill adults, and (3) which can be used to detect (risk of) the disorder at the individual level. This suggests that MRI-scans, after further improvement and independent validation, may be of added value in the risk profiling of BD and SZ.

Background: In recent decades, numerous in vivo brain imaging studies utilizing diffusion weighted MRI (dMRI) technique have focused on altered diffusivity in brains of patients with schizophrenia. However, the literature has not reached at consistent consensus despite a few interesting and promising results. In this study, we investigated whether or not various measures of dMRI (FA, AD, RD, and TR) are altered in patients with schizophrenia by comparing them in both patients and healthy controls with public neuroimaging data from SchizConnect (http://schizconnect. org). Methods: The final data set was consisted of 121 schizophrenia patients and 119 healthy controls. After verifying 161 anatomical regions of interest (ROIs), we estimated the mean value and standard deviation of fractional anisotropy (FA), axial diffusivity (AD), radial diffusivity (RD), and trace (TR) in each ROI among the healthy controls. After that, we calculated the Z-score of each single ROI in every individual brain of both patients and healthy controls. The Z-score information of each person is then integrated into two location-independent measures. One is the total number of "abnormal" lesions, in which the absolute Z-score is above the cut-off value estimated by the Bonferroni correction, and the other is the largest absolute Z-score. After all, by using Welch two-sample t-test, we compared these two measures between the groups of patients and healthy controls. Results: The number of abnormal lesions was notably increased in patients group, in terms of RD (p=0.01063) and TR (p=0.009329). Meanwhile, no statistically significant differences related to FA and AD were observed. On the other hand, it was found that the largest absolute Z-score was elevated in patients group, in terms of AD (p=0.03371), RD (p=0.0001762), and TR (p<0.00001). Otherwise, no significant differences related to FA were observed. Discussion: In this study, we found a few remarkable differences of familiar measures, especially TR, between brains of patients with schizophrenia and healthy controls. This suggests that there should be some subtle changes in the brains of patients with schizophrenia, including microstructural destruction.

UMC Utrecht
Background: Schizophrenia (SZ) and bipolar disorder (BD) are severe psychiatric disorders that are not easily distinguishable based on clinical measures. Offspring of patients with SZ or BD have a tenfold increased risk of developing the disorder as well as an increased risk for other severe mental disorders. Reliable identification of these subjects might allow for early recognition and intervention, which have been shown to be beneficial for treatment outcome and may even prevent transition to illness. Based on abundant evidence that SZ and BD are associated with structural brain abnormalities, we investigated whether MRI brain-scans can be used to detect individual risk of developing SZ or BD in adolescents. Methods: Structural MRI brain-scans were acquired in adolescent offspring (8-19 year) of parents with schizophrenia (oSZ;N=50), bipolar disorder (oBD;N=82), and without a mood or psychotic DSM-IV disorder (oHC;N=53), as part of the Dutch Bipolar and Schizophrenia Offspring Study (DBSOS). Support vector machine (SVM) models were trained on the gray matter tissue density maps to predict to which offspring class (oHC/oBD/oSZ) an individual belonged. Prediction accuracy was assessed using cross-validation. To validate our prediction models, we applied them to the tissue maps from subjects from a sample of unrelated HC/BD/SZ adults. Secondly, validated prediction models built from the adult subjects' MRI scans were applied to the tissue maps of the adolescents to predict illness class (HC/BD/SZ).

Results:
The offspring-based model separated oHC/oSZ individuals with 77% accuracy (p<0.001), oHC/oBD with 68% accuracy (p<0.001), and oBD/oSZ with 64% accuracy (p<0.01). The adult-based models could separate the patients' offspring from the healthy offspring with 66-70% accuracy, but oBD from oSZ with lower accuracy (59%). In addition, the offspring models could separate adult patients from control subjects with comparable accuracy (66-68%) and separate the two patient groups with moderate accuracy (69%). Discussion: The familial high-risk adolescents could be separated from controls with moderate to high accuracy (up to 77%), based on their MRIscans. Moreover, the brain tissue patterns based on risk (adolescents) or illness (adults) were able to predict (risk) class in the other stage group. These results show (1) that high-risk individuals already show brain abnormalities, and (2) display similarities with abnormalities in ill adults, and (3) which can be used to detect (risk of) the disorder at the individual level. This suggests that MRI-scans, after further improvement and independent validation, may be of added value in the risk profiling of BD and SZ.

F173. PITCH AND DURATION MISMATCH NEGATIVITY, AUDITORY CORTEX GRAY MATTER, AND PRODROMAL ROLE FUNCTIONING IN THE FIRST EPISODE SCHIZOPHRENIA SPECTRUM
Dean Salisbury* ,1 , Anna Shafer 2 , Brian Coffman 2 , Timothy Murphy 2 1 University of Pittsburgh School of Medicine; 2 University of Pittsburgh Background: Primary auditory cortex, contained within Heschl's gyrus, is implicated auditory processing deficits and auditory verbal hallucinations in schizophrenia. Previously we showed a pathological correlation between the magnitude of the pitch-deviant mismatch negativity (pMMN) response during a passive auditory task and reductions in gray matter volume in Heschl's gyrus in subjects with first hospitalized for schizophrenia. The aim of this study was to replicate this finding, examine duration-deviant mismatch negativity (dMMN) and gray matter correlations, and to examine pre-psychosis role functioning, in a first episode psychosis sample within the schizophrenia-spectrum. Methods: Participants included 40 first episode schizophrenia subjects (FESz) and 40 healthy controls (HC) matched for age, parental socioeconomic status, IQ, sex, and handedness. For MMN extracted from the EEG, standard tones were presented repeatedly (1 kHz, 75 dB, 50 ms pips, 5 ms rise/fall times, 330 ms SOA) with an occasional pitch deviant (1.2 kHz, 10% of trials) or duration deviant (100 ms, 10% of trials) interspersed. pMMN and dMMN were measured from subtraction waveforms as the average voltage within a 100-ms group averaged peak window at Fz. Role functioning was measured with the Cornblatt Global Functioning: Role scale. A subset of 28 FESz and 28 matched HC underwent structural MRI.
High-resolution T1-weighted structural MRI data (3T) were acquired for each subject. Freesurfer was used to segment white matter, gray matter, and pial surfaces. Left and right Heschl's gyri were manually edited regions of interest, and gray matter volumes determined. Results: Despite a lack of pMMN or dMMN reduction at the group level in FESz, both measures were pathologically correlated with role functioning in the year prior to hospitalization. In FESz, smaller pMMN at Fz was associated with poorer role functioning in the year prior to psychosis (rho= -.35, p =.03). Similar associations were observed for dMMN (rho= -.41, p <.01). Furthermore, in the subset of FESz with sMRI, smaller pMMN at Fz was associated with less total gray matter volume in left Heschl's gyrus (TGMV) (rho= -.40, p =.03) but not right. Similar associations were observed for dMMN (rho= -.47, p .01). As well, role functioning and auditory cortex gray matter volumes were not correlated in FESz. There were no significant correlations within HC. Discussion: Although pMMN and dMMN are not reduced at the group level, the size of both are associated with impaired functioning prior to psychosis and reduced gray matter volume of left hemisphere Heschl's gyrus, containing primary and secondary auditory cortices. Thus, pMMN and dMMN although not sufficient as biomarkers of disease presence, are suitable as reliable biomarkers of disease progression. Presumably, poorer role functioning and less gray matter reflect more of the pre-psychosis progressive pathological process thought to occur in the prodromal phase of psychosis. Hence, pMMN and dMMN are likely to serve as sensitive and robust outcome measures for therapeutic interventions and to guide treatment strategies in the prodrome and during early psychosis.

Hospital Presidente Vargas -Brazil
Background: Obesity is associated with both structural and functional changes of the central nervous system, and is frequent in psychiatry settings. The increased prevalence of obesity in schizophrenia (SCZ) and bipolar disorder (BD) is associated with illness severity, functioning impairment and cognitive deficits. It cannot be attributed to biases inherent in treatment-seeking samples, given that this association is detectable even in drugnaïve patients. Diffusion tensor imaging (DTI) analyses of major brain fibers in both disorders show shared abnormalities of white matter. DTI has been employed as a highly sensitive tool to investigate microstructural changes in white matter structure. While gray matter alterations in obesity point to a consistent reduction with increasing body mass index (BMI), volumetric changes in white matter are more complex and less conclusive. Fractional anisotropy (FA) is the most commonly used parameter as it is the best estimate of fiber integrity as well as axonal and myelin degeneration, and has been reported an association with BMI in depressed BD patients, but not explored in SCZ nor in comparison with a control group (CTR). The aim of this study was to analyze the relationship between obesity and brain alterations assessed by DTI in SCZ, BD and CTR. Methods: In one-hundred fifty (N=150) individuals (SCZ:49; BD:35; CTR:66) were administered clinical rating scales, collected sociodemographic data and submitted to magnetic resonance imaging (MRI) acquisition in a 1.5 T machine. Linear regression models were performed independently for each group in order to test the relationship of BMI on