1. Deep Learning for Quality Control of Subcortical Brain 3D Shape Models
- Author
-
Henrik Walter, Anita Richter-Rossler, Elliot Hong, Vince D. Calhoun, Pedro G.P. Rosa, Matthew D. Sacchet, Gianfranco Spalletta, Stefan Ehrlich, Ole A. Andreassen, Jair C. Soarez, Mon-Ju Wu, Martin Walter, Peter Kochunov, Aristotle Voineskos, Raymond Salvador, Kathryn I. Alpert, Dick J. Veltman, Dan J. Stein, Alexander J. Huang, Udo Dannlowski, Edith Pomarol-Clotet, Ilya M. Veer, Gary Donohoe, Steven G. Potkin, Artemis Zavaliangos-Petropolu, Stefan Borgwardt, Theodore D. Satterthwaite, Li Shen, Ruben C. Gur, Shan Cong, Andr Aleman, Erin W. Dickie, Nerisa Banaj, Simon Cervenka, Ingrid Agartz, Paola Fuentes-Claramonte, Dmitry Petrov, Jessica A. Turner, Meng Li, Fabrizio Piras, Bernhard T. Baune, Paul M. Thompson, Roberto Roiz-Santiaez, Nynke A. Groenewold, Diana Tordesillas-Gutirrez, Alexander Tomyshev, Daniela Vecchio, Mauricio H. Serpa, Nhat Trung Doan, Anthony A. James, Lei Wang, Fleur M. Howells, Geraldo F. Busatto, Lianne Schmaal, Christopher R.K. Ching, Anne Uhlmann, Theo G.M. van Erp, Irina V. Lebedeva, Marcus V. Zanetti, Boris A. Gutman, Erick J. Canales-Rodríguez, Sinead Kelly, Benson Irungu, Dmitry Isaev, Benedicto Crespo-Favorro, David C. Glahn, Joaquim Radua, Lars T. Westlye, Dominik Grotegerd, Fabienne Harrisberger, Valentina Ciullo, Esther Walton, Ian H. Gotlib, and Egor Kuznetsov
- Subjects
Artificial neural network ,business.industry ,Computer science ,Deep learning ,Schizophrenia (object-oriented programming) ,Pattern recognition ,Grey matter ,medicine.disease ,Residual neural network ,medicine.anatomical_structure ,Schizophrenia ,Feature (computer vision) ,medicine ,Major depressive disorder ,Artificial intelligence ,business ,Parametric statistics - Abstract
We present several deep learning models for assessing the morphometric fidelity of deep grey matter region models extracted from brain MRI. We test three different convolutional neural net architectures (VGGNet, ResNet and Inception) over 2D maps of geometric features. Further, we present a novel geometry feature augmentation technique based on parametric spherical mapping. Finally, we present an approach for model decision visualization, allowing human raters to see the areas of subcortical shapes most likely to be deemed of failing quality by the machine. Our training data is comprised of 5200 subjects from the ENIGMA Schizophrenia MRI cohorts, and our test dataset contains 1500 subjects from the ENIGMA Major Depressive Disorder cohorts. Our final models reduce human rater time by 46-70%. ResNet outperforms VGGNet and Inception for all of our predictive tasks.
- Published
- 2018
- Full Text
- View/download PDF