Machine learning techniques are poised to become clinically useful methods that may be used for diagnosis, prognosis, and treatment decisions. Despite this, they are currently underutilised in medical studies and, even more in psychiatric research because most current tools require strong programming and computational engineering skills (e.g., scikit-learn, caret, Weka, nilearn). While there are some great tools that do not require programming experience (e.g., PRoNTo), these are often focused on making predictions from specific data domains such as neuroimaging data. This highlights a pressing need for user-friendly machine learning software that makes advanced methods available to clinical researchers from different fields aiming at collaboratively developing diagnostic, predictive, and prognostic tools for precision medicine approaches.
NeuroMiner has been continuously developed by Nikolaos Koutsouleris since 2009 (with support from the European Commission for the PRONIA project since 2013) to provide clinical researchers with cutting-edge machine learning methods for the analysis of heterogeneous data domains, such as clinical and neurocognitive read-outs, structural and functional neuroimaging data, and genetic information. The program can be considered as an interface to a large variety of unsupervised, and supervised pattern recognition algorithms that have been developed in the machine learning field over the last decades. Furthermore, the application implements itself different strategies for preprocessing, filtering and fusing heterogeneous data, training ensembles of predictors and visualizing and testing the significance of the computed predictive patterns. The current release candidate of NeuroMiner has been tested in the Section of Neurodiagnostic Applications on a variety of datasets from healthy controls and patients with psychiatric disorders and was designed specifically to create robust models with a high probability of generalization to new datasets. For reference, we include here a list of papers, which were all based on previous versions of the program.
More specifically, using a light-weight and interactive text-based menu system, NeuroMiner allows the user to:
- load their data easily (e.g., using spreadsheets, NifTi images, or SPM structures);
- build a variety of cross-validation frameworks for classification and regression problems that have become a gold standard in the field (e.g., repeated nested cross-validation, leave-site-out cross-validation);
- apply a range of preprocessing strategies (e.g., scaling, filtering, many forms of dimensionality reduction, etc.);
- choose and combine cutting-edge supervised algorithms (e.g., support vector machine, elastic net, random forest, etc.);
- apply feature selection procedures (e.g., wrappers), data fusion techniques, and stacked generalization;
- apply learned models to new data (external validation).
To assist in selecting and analysing data, the user can visualise the data during input, monitor accuracy during learning, and understand the results of complex analyses using multiple display options. These allow the user to accurately report the data and also to understand the underlying machine learning analyses. Furthermore, the ability to apply the learned models to completely new data is important because it is quickly becoming a standard requirement of all machine learning studies. Combined, NeuroMiner gives the user the opportunity to design, implement, understand, and report machine learning analyses.
Please note that NeuroMiner is copyright software ((c) 2017, all rights reserved) and at this stage no distribution (commercial or non-commercial) is permitted because it is still being tested. NeuroMiner is supplied as is and no formal maintenance is provided or implied. In no event shall the copyright holder be liable to any party for direct, indirect, special, incidental, or consequential damages, including lost profits, arising out of the use of this software and its documentation, even if the copyright holder has been advised of the possibility of such damage. The copyright holder specifically disclaims any warranties, including, but not limited to, the implied warranties of merchantability and fitness for a particular purpose. The software and accompanying documentation provided hereunder is provided “as is”. The copyright holder has no obligation to provide maintenance, support, updates, enhancements, or modifications (but we plan to).
It is currently release-candidate software (NM 0.992) that is undergoing regular monthly updates. The release of NeuroMiner 1.0 is planned for July 2017. Please send us any comments, questions, or bug reports by email.
Papers that used NM:
1. Koutsouleris N, Meisenzahl EM, Davatzikos C, Bottlender R, Frodl T, Scheuerecker J, Schmitt G, Zetzsche T, Decker P, Reiser M, Möller HJ, Gaser C. Use of neuroanatomical pattern classification to identify subjects in at-risk mental states of psychosis and predict disease transition. Archives of General Psychiatry. 2009; 66(7):700-12
2. Koutsouleris N, Meisenzahl EM, Borgwardt S, Riecher-Rössler A, Frodl T, Kambeitz J, Köhler Y, Falkai P, Möller H.-J., Reiser M, Davatzikos C. Individualized differential diagnosis of schizophrenia and mood disorders using neuroanatomical biomarkers. Brain. 2015 Jul;138(Pt 7):2059-73.
3. Koutsouleris N, Kahn RS, Chekroud AM, Leucht S, Falkai P, Wobrock T, Derks EM,Fleischhacker WW, Hasan A. Multisite prediction of 4-week and 52-week treatment outcomes in patients with first-episode psychosis: a machine learning approach. Lancet Psychiatry. 2016 Oct;3(10):935-946. doi: 10.1016/S2215-0366(16)30171-7.
4. Cabral C, Kambeitz-Ilankovic L, Kambeitz J, Calhoun VD, Dwyer DB, von Saldern S, Urquijo MF, Falkai P, Koutsouleris N. Classifying Schizophrenia Using Multimodal Multivariate Pattern Recognition Analysis: Evaluating the Impact of Individual Clinical Profiles on the Neurodiagnostic Performance. Schizophrenia Buletinl. 2016 Jul;42 Suppl 1:S110-7. doi: 10.1093/schbul/sbw053.
5. Koutsouleris N, Borgwardt S, Meisenzahl EM, Bottlender R, Möller HJ, Riecher-Rössler A. Disease prediction in the at-risk mental state for psychosis using neuroanatomical biomarkers: results from the FePsy-study. Schizophrenia Bulletin. 2012; 38(6):1234-46
6. Borgwardt SJ, Koutsouleris N, Aston J, Studerus E, Smieskova R, Riecher-Rössler A, Meisenzahl EM. Distinguishing prodromal from first-episode psychosis using neuroanatomical pattern recognition: Evidence from single-subject structural MRI. Schizophrenia Bulletin. 2013; 39(5):1105-14. doi: 10.1093/schbul/sbs095
7. Koutsouleris N, Davatzikos C, Bottlender R, Patschurek-Kliche K, Scheuerecker J, Decker P, Gaser C, Möller HJ; Meisenzahl E. Early recognition and disease prediction in the at-risk mental states for psychosis using neurocognitive pattern classification. Schizophrenia Bulletin. 2012; 38(6):1200-15
8. Koutsouleris N, Riecher-Rössler A, Meisenzahl E, Smieskova R, Studerus E, Kambeitz-Ilankovic L, von Saldern S, Cabral C, Reiser M, Falkai P, Borgwardt S. Detecting the psychosis prodrome across high-risk populations using neuroanatomical biomarkers. Schizophrenia Bulletin. 2014, 41(2):471-82.
9. Koutsouleris N, Davatzikos C, Borgwardt S, Gaser C, Bottlender R, Frodl T, Falkai P, Riecher-Rössler A, Möller HJ, Reiser M, Pantelis C, Meisenzahl E. Accelerated Brain Aging in Schizophrenia and Beyond: A Neuroanatomical Marker of Psychiatric Disorders. Schizophrenia Bulletin. 2014 Sep;40(5):1140-53
10. Koutsouleris N, Gaser C, Bottlender R, Davatzikos C, Decker P, Jäger M, Schmitt G, Reiser M, Möller HJ, Meisenzahl EM, Use of Neuroanatomical Pattern Regression to Predict the Structural Brain Dynamics of Vulnerability and Transition to Psychosis. Schizophrenia Research. 2010;123(2-3):175-187
11. Kambeitz-Ilankovic L, Meisenzahl EM, Cabral C, von Saldern S, Kambeitz J, Falkai P, Möller HJ, Reiser M, Koutsouleris N. Prediction of outcome in the psychosis prodrome using neuroanatomical pattern classification. Schizophrenia Research. 2015;173(3):159-65.