Improved individualized prediction of schizophrenia in subjects at familial high risk, based on neuroanatomical data, schizotypal and neurocognitive features Eleni Zarogianni 1 , Amos J. Storkey 2 , Eve C. Johnstone 1 , David G. C. Owens 1 , Stephen M. Lawrie 1 1 Division of Psychiatry, School of Clinical Sciences, University of Edinburgh, the Royal Edinburgh Hospital, Morningside Park, UK 2 Institute for Adaptive and Neural Computation, University of Edinburgh, UK Correspondence to: Eleni Zarogianni, Kennedy Tower, Division of Psychiatry, School of Clinical Sciences, University of Edinburgh, the Royal Edinburgh Hospital, Morningside Park, EH10 5HF, UK. E-mail: [email protected]. Tel.: +44(0) 131 537 6182. Highlights: The individualized prediction of schizophrenia is possible in familial high-risk cohort by using baseline neuroanatomical data. Combination of neuroanatomical data with schizotypal and neurocognitive measures increased the diagnostic accuracy of the classifier. Keywords: MRI, machine learning, Support Vector Machine, Recursive Feature Elimination, schizophrenia, prediction, familial HR Abstract 1
38
Embed
€¦ · Web viewImproved individualized prediction of schizophrenia in subjects at familial high risk, based on ... World Health Organization ... Intelligence Quotient; RISC, ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Improved individualized prediction of schizophrenia in subjects at familial high risk, based on neuroanatomical data, schizotypal and neurocognitive features
Eleni Zarogianni1, Amos J. Storkey2, Eve C. Johnstone1, David G. C. Owens1, Stephen M. Lawrie1
1Division of Psychiatry, School of Clinical Sciences, University of Edinburgh, the Royal Edinburgh Hospital, Morningside Park, UK2Institute for Adaptive and Neural Computation, University of Edinburgh, UK
Correspondence to: Eleni Zarogianni, Kennedy Tower, Division of Psychiatry, School of Clinical Sciences, University of Edinburgh, the Royal Edinburgh Hospital, Morningside Park, EH10 5HF, UK. E-mail: [email protected]. Tel.: +44(0) 131 537 6182.
Highlights: The individualized prediction of schizophrenia is possible in familial
high-risk cohort by using baseline neuroanatomical data. Combination of
neuroanatomical data with schizotypal and neurocognitive measures increased the
diagnostic accuracy of the classifier.
Keywords: MRI, machine learning, Support Vector Machine, Recursive Feature
Bois C, Whalley HC, McIntosh AM, et al., 2015. Structural magnetic resonance imaging markers of susceptibility and transition to schizophrenia: a review of familial and clinical high risk population studies. J. Psychopharmacol, 29(2):144-154.
Borgwardt SJ, McGuire PK, Aston Jet al., 2007a. Structural brain abnormalities in individuals with an at-risk mental state who later develop psychosis. Br J Psychiatry Suppl, 51:s69-75. doi:10.1192/bjp.191.51.s69.
Borgwardt SJ, Riecher-Rössler A, Dazzan P, et al., 2007b. Regional gray matter volume abnormalities in the at risk mental state. Biol. Psychiatry. 61(10):1148-1156. doi:10.1016/j.biopsych.2006.08.009.
Borgwardt SJ, McGuire PK, Aston J, et al., 2008. Reductions in frontal, temporal and parietal volume associated with the onset of psychosis. Schizophrenia Res. 106(2–3):108-114..
Borgwardt S, McGuire P, Fusar-Poli P. 2011. Gray matters!--mapping the transition to psychosis. Schizophr Res.133(1-3):63-67. doi:10.1016/j.schres.2011.08.021.
Brewer WJ, Wood SJ, Phillips LJ, et al. 2006. Generalized and specific cognitive performance in clinical high-risk cohorts: a review highlighting potential vulnerability markers for psychosis. Schizophr Bull. 32(3):538-555. doi:10.1093/schbul/sbj077.
Burges C. 1998 A tutorial on Support Vector Machines for Pattern Recognition. Data Mining and Knowledge Discovery. Volume 2, Issue 2, pp 121-167.
Byrne M, Clafferty BA, Cosway R, et al. 2003. Neuropsychology, genetic liability, and psychotic symptoms in those at high risk of schizophrenia. J Abnorm Psychol. 112(1):38-48.
Cannon, T.D., Cadenhead, K., Cornblatt, B., et al. 2008. Prediction of psychosis in youth at high clinical risk: a multisite longitudinal study in North America. Arch. Gen. Psychiatry. 65, 28–37. doi:10.1001/archgenpsychiatry.2007.3
Chang CC and Lin CJ. 2011. LIBSVM : a library for support vector machines. ACM Transactions on Intelligent Systems and Technology. 2:27:1--27:27.
Cooper D, Barker V, Radua J, et al., 2014. Multimodal voxel-based meta-analysis of structural and functional magnetic resonance imaging studies in those at elevated genetic risk of developing schizophrenia. Psychiatry Res. 221(1):69-77.
Davatzikos C, Shen D, Gur RC, et al., 2005. Whole-brain morphometric study of schizophrenia revealing a spatially complex set of focal abnormalities. Arch. Gen. Psychiatry, 62(11):1218-1227. doi:10.1001/archpsyc.62.11.1218.
De Martino F, Valente G, Staeren N, et al., 2008. Combining multivariate voxel selection and support vector machines for mapping and classification of fMRI spatial patterns. Neuroimage. 43(1):44-58. doi:10.1016/j.neuroimage.2008.06.037.
Demjaha, A., Valmaggia, L., Stahl, D., et al., 2012. Disorganization/cognitive and negative symptom dimensions in the at-risk mental state predict subsequent transition to psychosis. Schizophr Bull 38, 351–359. doi:10.1093/schbul/sbq088
Dubey R, Zhou J, Wang Y, Thompson PM, Ye J., 2014. Analysis of sampling techniques for imbalanced data: an N=648 ADNI Study. NeuroImage. 87:220-241.
Duchesnay E, Cachia A, Boddaert Net al., 2011. Feature selection and classification of imbalanced datasets: application to PET images of children with autistic spectrum disorders. Neuroimage. 57(3): 1003-14.
Erlenmeyer-Kimling L, Rock D, Roberts SA, et al., 2000. Attention, memory, and motor skills as childhood predictors of schizophrenia-related psychoses: the New York High-Risk Project. Am J Psychiatry. 157(9):1416-1422.
Fan Y, Gur RE, Gur RC, et al., 2008. Unaffected family members and schizophrenia patients share brain structure patterns: a high-dimensional pattern classification study. Biol. Psychiatry. 63(1):118-124.
Fusar-Poli P, Borgwardt S, Crescini A, et al., 2011. Neuroanatomy of vulnerability to psychosis: a voxel-based meta-analysis. Neurosci Biobehav Rev. 35(5):1175-1185.
Fusar-Poli P, Deste G, Smieskova R, et al. 2012. Cognitive functioning in prodromal psychosis: a meta-analysis. Arch Gen Psychiatry. 69(6):562-571.
Fusar-Poli P, Borgwardt S, Bechdolf A, et al. 2013. The psychosis high-risk state: a comprehensive state-of-the-art review. JAMA Psychiatry. 70(1):107-120.
Guyon I, Weston J, Barnhill S, Vapnik V. 2002. Gene Selection for Cancer Classification Using Support Vector Machines. Mach Learn. 46(1-3):389–422.
Haijma, S. V., Van Haren, N., Cahn, W., et al., 2013. Brain volumes in schizophrenia: a meta-analysis in over 18 000 subjects. Schizophrenia Bull. 39(5), 1129–1138.
He H and Garcia EA. 2009. Learning from Imbalanced Data. IEEE Transactions on Knowledge and Data engineering. Vol 21. No9.
Heinrichs RW, Zakzanis KK. 1998. Neurocognitive deficit in schizophrenia: a 15
quantitative review of the evidence. Neuropsychology. 12(3):426-445.
HMSO, 1991. Classification of Occupations of the Registrar General. HMSO, London.
Hodges A, Byrne M, Grant E, Johnstone E. 1999. People at risk of schizophrenia. Sample characteristics of the first 100 cases in the Edinburgh High-Risk Study. Br J Psychiatry. 174:547-553.
Honea R, Crow TJ, Passingham D, Mackay CE. 2005. Regional deficits in brain volume in schizophrenia: a meta-analysis of voxel-based morphometry studies. Am J Psych. 162(12):2233–2245. doi: 10.1176/appi.ajp.162.12.2233.
Jablensky A. 1997. The 100-year epidemiology of schizophrenia. Schizophr. Res. 28(2-3):111-125.
Japkowicz N, Stephen S. 2002. The class imbalance problem: a systematic study. Journal of Intelligent Data Analysis, 5 (6):429-449
Job DE, Whalley HC, McConnell S, et al., 2003. Voxel-based morphometry of grey matter densities in subjects at high risk of schizophrenia. Schizophr Res. 64(1):1-13.
Job DE, Whalley HC, McIntosh AM, et al., 2006. Grey matter changes can improve the prediction of schizophrenia in subjects at high risk. BMC Med. 4:29. doi:10.1186/1741-7015-4-29.
Johnstone EC, Abukmeil SS, Byrne M, et al., 2000.Edinburgh high risk study--findings after four years: demographic, attainment and psychopathological issues. Schizophr Res. 46(1):1-15.
Johnstone EC, Ebmeier KP, Miller P, et al., 2005. Predicting schizophrenia: findings from the Edinburgh High-Risk Study. Br J Psychiatry. 186:18-25. doi:10.1192/bjp.186.1.18.
Karageorgiou E, Schulz SC, Gollub RL, et al., 2011 Neuropsychological testing and structural magnetic resonance imaging as diagnostic biomarkers early in the course of schizophrenia and related psychoses. Neuroinformatics. 9(4):321-333.
Klauser, P., Zhou, J., Lim, J.K.W., et al. 2015. Lack of Evidence for Regional Brain Volume or Cortical Thickness Abnormalities in Youths at Clinical High Risk for Psychosis: Findings From the Longitudinal Youth at Risk Study. Schizophr Bull 41, 1285–1293.
Klöppel S, Abdulkadir A, Jack CR, et al., 2012. Diagnostic neuroimaging across diseases. Neuroimage. 61(2):457-463. doi:10.1016/j.neuroimage.2011.11.002.
16
Koutsouleris N, Meisenzahl EM, Davatzikos C, et al., 2009a. Use of neuroanatomical pattern classification to identify subjects in at-risk mental states of psychosis and predict disease transition. Arch Gen Psychiatry. 66(7):700-712.
Koutsouleris N, Schmitt GJE, Gaser C, et al., 2009b. Neuroanatomical correlates of different vulnerability states for psychosis and their clinical outcomes. Br J Psychiatry. 2009b;195(3):218-226.
Koutsouleris N, Borgwardt S, Meisenzahl EMet al., 2012a. Disease prediction in the at-risk mental state for psychosis using neuroanatomical biomarkers: results from the FePsy study. Schizophr Bull. 38(6):1234-1246. doi:10.1093/schbul/sbr145.
Koutsouleris N, Davatzikos C, Bottlender R, et al., 2012b. Early recognition and disease prediction in the at-risk mental states for psychosis using neurocognitive pattern classification. Schizophr Bull. 38(6):1200-1215. doi:10.1093/schbul/sbr037.
Koutsouleris N, Riecher-Rössler A, Meisenzahl EM, et al., 2015. Detecting the psychosis prodrome across high-risk populations using neuroanatomical biomarkers. Schizophr Bull. 2015;41(2):471-482.
Lawrie SM, Whalley HC, Abukmeil SS, et al., 2001. Brain structure, genetic liability, and psychotic symptoms in subjects at high risk of developing schizophrenia. Biol Psychiatry. 49(10):811-823.
Lencz T, Smith CW, McLaughlin D, et al., 2006. Generalized and specific neurocognitive deficits in prodromal schizophrenia. Biol Psychiatry. 59(9):863-871.
Lymer GKS, Job DE, William T, et al., 2006. Brain-behaviour relationships in people at high genetic risk of schizophrenia. Neuroimage. 33(1):275-285..
Marshall M, Lockwood A. 2004. Early Intervention for psychosis. Cochrane Database Syst Rev (2):CD004718. doi:10.1002/14651858.CD004718.
Meisenzahl EM, Koutsouleris N, Gaser C, et al., 2008. Structural brain alterations in subjects at high-risk of psychosis: a voxel-based morphometric study. Schizophr Res. 102(1-3):150-162.
Miller PM, Lawrie SM, Byrne Met al., 2002. Self-rated schizotypal cognitions, psychotic symptoms and the onset of schizophrenia in young people at high risk of schizophrenia. Acta Psychiatr Scand. 105(5):341-345.
Mourão-Miranda J, Oliveira L, Ladouceur CD, et al., 2012a. Pattern recognition and functional neuroimaging help to discriminate healthy adolescents at risk for mood disorders from low risk adolescents. PLoS ONE. 7(2):e29482. doi:10.1371/journal.pone.0029482.
17
Mourao-Miranda J, Reinders A a. TS, Rocha-Rego V, et al., 2012b. Individualized prediction of illness course at the first psychotic episode: a support vector machine MRI study. Psychol Med. 42(5):1037-1047. doi:10.1017/S0033291711002005
Nieuwenhuis M, van Haren NEM, Hulshoff Pol HE, et al., 2012. Classification of schizophrenia patients and healthy controls from structural MRI scans in two large independent samples. Neuroimage. 61(3):606-612. doi:10.1016/j.neuroimage.2012.03.079.
Orrù G, Pettersson-Yeo W, Marquand AF, et al., 2012. Using Support Vector Machine to identify imaging biomarkers of neurological and psychiatric disease: a critical review. Neurosci Biobehav Rev. 36(4):1140-1152. doi:10.1016/j.neubiorev.2012.01.004.
Pantelis C, Velakoulis D, McGorry PD, et al., 2003. Neuroanatomical abnormalities before and after onset of psychosis: a cross-sectional and longitudinal MRI comparison. Lancet.361(9354):281-288. doi:10.1016/S0140-6736(03)12323-9.
Pantelis VC, Yücel M, Bora E, et al., 2009. Neurobiological markers of illness onset in psychosis and schizophrenia: The search for a moving target. Neuropsychol Rev.19:385–398.
Pettersson-Yeo W, Benetti S, Marquand AF, et al. 2013. Using genetic, cognitive and multi-modal neuroimaging data to identify ultra-high-risk and first-episode psychosis at the individual level. Psychol Med. 43(12):2547-2562. doi:10.1017/S003329171300024X.
Pettersson-Yeo, W., Benetti, S., Marquand, A.F., et al., 2014. An empirical comparison of different approaches for combining multimodal neuroimaging data with support vector machine. Front. Neurosci 8, 189. doi:10.3389/fnins.2014.00189
Pukrop R, Ruhrmann S, Schultze-Lutter F, Bechdolf A, et al. 2007. Neurocognitive indicators for a conversion to psychosis: comparison of patients in a potentially initial prodromal state who did or did not convert to a psychosis. Schizophr Res. 2007;92(1-3):116-125.
Rey, A., 1964. L'examen clinique en psychologie. Presses Universitaires de France, Paris.
Riecher-Rössler A, Gschwandtner U, Borgwardt S, et al., 2006. Early detection and treatment of schizophrenia: how early? Acta Psychiatr Scand Suppl.(429):73-80.
Riecher-Rössler A, Pflueger MO, Aston J, et al., 2009. Efficacy of using cognitive status in predicting psychosis: a 7-year follow-up. Biol Psychiatry. 66(11):1023-1030.
Ruhrmann S, Schultze-Lutter F, Salokangas RK, et al., 2010. Prediction of psychosis in adolescents and young adults at high risk: results from the prospective European
18
Prediction of Psychosis Study. Arch Gen Psychiatry. 67:241–251.
Rust J. 1988. The Rust Inventory of Schizotypal Cognitions (RISC). Schizophr Bull. 14(2):317-322.
Schnack HG, Nieuwenhuis M, van Haren NEM, et al., 2014. Can structural MRI aid in clinical classification? A machine learning study in two independent samples of patients with schizophrenia, bipolar disorder and healthy subjects. Neuroimage. 84:299-306.
Seidman, L.J., Giuliano, A.J., Meyer, E.C., et al., 2010. Neuropsychology of the prodrome to psychosis in the NAPLS consortium: relationship to family history and conversion to psychosis. Arch. Gen. Psychiatry. 67, 578–588. doi:10.1001/archgenpsychiatry.2010.66
Smieskova R, Fusar-Poli P, Allen P, et al., 2010. Neuroimaging predictors of transition to psychosis--a systematic review and meta-analysis. Neurosci Biobehav Rev. 34(8):1207-1222.
Sun D, van Erp TGM, Thompson PM, et al., 2009. Elucidating a magnetic resonance imaging-based neuroanatomic biomarker for psychosis: classification analysis using probabilistic brain atlas and machine learning algorithms. Biol. Psychiatry. 66(11):1055-
Tzourio-Mazoyer N, Landeau B, Papathanassiou D, et al., 2002. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage. 15(1):273-289. doi:10.1006/nimg.2001.0978.
Van Hulse J, Khoshgoftaar TM, Napolitano A. 2007. Experimental Perspectives on Learning from Imbalanced Data Proceedings of the 24th International Conference on Machine Learning. ACM, Corvalis, Oregon , pp. 935–942
Vapnik V. 2000. The Nature of Statistical Learning Theory. Springer.
Wei Q, Dunbrack RL. 2013.The Role of Balanced Training and testing data Sets for Binary Classifiers in Bioinformatics. Plos ONE. 8(7).
Wing JK, Cooper JE, Sartorius N. 1974. Measurement and Classification of Psychiatric Symptoms, an Instruction Manual for the PSE and CATEGO Programme. London: Cambridge University Press.
Whyte M-C, Brett C, Harrison LK, et al., 2006. Neuropsychological performance over time in people at high risk of developing schizophrenia and controls. Biol Psychiatry. 59(8):730-739.
19
Wood, S.J., Yücel, M., Velakoulis, D., et al., 2005. Hippocampal and anterior cingulate morphology in subjects at ultra-high-risk for psychosis: the role of family history of psychotic illness. Schizophr. Res. 75, 295–301.
World Health Organization. 1994. International classification of diseases, 10th edn. Geneva: WHO.
Wright IC, Rabe-Hesketh S, Woodruff PWet al., 2000. Meta-analysis of regional brain volumes in schizophrenia. Am J Psych. 157(1):16–25. doi: 10.1176/ajp.157.1.16.
Yang H, Liu J, Sui J, et al., 2010. A Hybrid Machine Learning Method for Fusing fMRI and Genetic Data: Combining both Improves Classification of Schizophrenia. Front Hum Neurosci. 4:192. doi:10.3389/fnhum.2010.00192.
Zarogianni E, Moorhead TWJ, Lawrie SM. 2013. Towards the identification of imaging biomarkers in schizophrenia, using multivariate pattern classification at a single-subject level. Neuroimage Clin. 3:279-289.
20
Figure 1. Representation of a linear, binary SVM classifier. (a) Illustration of the classification problem between two groups (i.e. circles represent patients, squares represent healthy controls) for the simplified case of two voxels. The dashed lines represent a subset of possible separating hyperplanes, described by a weight vector w and an offset b. (b) the optimal hyperplane is the one with the largest margin of separation between the groups. The weight vector w is perpendicular to the separating surface and points to the direction of maximum discrimination. The circled points represent support vectors.
Figure 2. Representation of the nested LOO-CV SMV-RFE method. We employed a nested LOO-CV where we repeatedly excluded one subject to comprise the testing set and the remaining subjects were again repeatedly repartitioned in an internal validation loop where one subject was left out for validation and the rest formed the internal training group. In this loop, RFE was repeatedly performed and the mean accuracy on the validation group at each elimination level was recorded until all features were removed. The feature set that produced the maximum accuracy on the validation set was selected and applied to the testing set of the outer testing loop. Finally, mean accuracy was calculated across all outer CV loops. It should be noted that the nested LOO-CV provided an unbiased estimate of the expected diagnostic accuracy on new cases.
Figure 3. Discrimination map for the classification of HR[ill] vs HR[symp]: a) just baseline MRI data were considered and b) baseline MRI were combined with RISC and RAVLT variables. The colours represent the weight of each feature in the classification function (the red scale represents positive weights and the blue scale represents negative weights). The SVM weight vector is a linear combination or weighted average of the support vectors and defines the decision boundary. The weight vector is therefore a spatial representation of the decision boundary. Every feature contributes with a certain weight to the decision boundary or classification function. Given a positive and a negative class (+1=HR[ill]; -1=HR[symp] group), a positive weight means the weighted average in that region was higher for the HR[ill] group, and a negative weight means the weighted average was higher for HR[symp] group. Note: features correspond to GM volume measures in the AAL-defined brain regions, and not voxels.
21
Table 1. Baseline sociodemograpchic and behavioural assessment variables of the groups
HR[ill] HR[symp] P HR[symp]test
Number of participants 17 17 40
Mean age at baseline (SD) 20.07(2.37) 20.03(2.6) nsa 21.07(4.59)
Male (%) 11(64.7) 11(64.7) nsb 14(35)
Mean RISC score (SD) 39.88(10.6) 25.23(11.75) <0.01a 33.45(13.39)
Mean RAVLT, trials 1-5 (SD) 47.64(7.49) 53.41(7.45) nsa ** 51.15(11.59)
Mean WAIS-IQ (SD) 98.64(12.93) 98.98(14.7) nsa 96.42(12.64)
HR[ill]: individuals at high familial risk who developed schizophrenia during follow-up period; HR[symp]: individuals at high familial risk who remained well but developed psychotic symptoms during follow-up period; IQ, Intelligence Quotient; RISC, Rust Inventory of Schizotypal Cognitions; RAVLT, Rey Auditory Verbal Learning Test; WAIS-R, Wechsler Adult Intelligence Scale- Revised. Social class of origin was based on the father's occupation at the time of the subject's birth using the Occupational Classification of the Registrar General (HMSO, 19991).
Statistical analyses are shown for the HR[ill]-HR[symp] contrast. b Fisher’s exact tests were applied to sex, handedness, social class of origin, present cannabis use, smoking and symptom severity scores. a
All other variables were evaluated using T tests, P significance. **effect size was r=0.54
Table 2. Diagnostic performance of the classifier, using the sMRI data only, and sMRI data combined with baseline behavioural variables