1
|
Bourached A, Bonkhoff AK, Schirmer MD, Regenhardt RW, Bretzner M, Hong S, Dalca AV, Giese AK, Winzeck S, Jern C, Lindgren AG, Maguire J, Wu O, Rhee J, Kimchi EY, Rost NS. Scaling behaviours of deep learning and linear algorithms for the prediction of stroke severity. Brain Commun 2024; 6:fcae007. [PMID: 38274570 PMCID: PMC10808016 DOI: 10.1093/braincomms/fcae007] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Key Words] [Grants] [Track Full Text] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/14/2023] [Revised: 09/01/2023] [Accepted: 01/09/2024] [Indexed: 01/27/2024] Open
Abstract
Deep learning has allowed for remarkable progress in many medical scenarios. Deep learning prediction models often require 105-107 examples. It is currently unknown whether deep learning can also enhance predictions of symptoms post-stroke in real-world samples of stroke patients that are often several magnitudes smaller. Such stroke outcome predictions however could be particularly instrumental in guiding acute clinical and rehabilitation care decisions. We here compared the capacities of classically used linear and novel deep learning algorithms in their prediction of stroke severity. Our analyses relied on a total of 1430 patients assembled from the MRI-Genetics Interface Exploration collaboration and a Massachusetts General Hospital-based study. The outcome of interest was National Institutes of Health Stroke Scale-based stroke severity in the acute phase after ischaemic stroke onset, which we predict by means of MRI-derived lesion location. We automatically derived lesion segmentations from diffusion-weighted clinical MRI scans, performed spatial normalization and included a principal component analysis step, retaining 95% of the variance of the original data. We then repeatedly separated a train, validation and test set to investigate the effects of sample size; we subsampled the train set to 100, 300 and 900 and trained the algorithms to predict the stroke severity score for each sample size with regularized linear regression and an eight-layered neural network. We selected hyperparameters on the validation set. We evaluated model performance based on the explained variance (R2) in the test set. While linear regression performed significantly better for a sample size of 100 patients, deep learning started to significantly outperform linear regression when trained on 900 patients. Average prediction performance improved by ∼20% when increasing the sample size 9× [maximum for 100 patients: 0.279 ± 0.005 (R2, 95% confidence interval), 900 patients: 0.337 ± 0.006]. In summary, for sample sizes of 900 patients, deep learning showed a higher prediction performance than typically employed linear methods. These findings suggest the existence of non-linear relationships between lesion location and stroke severity that can be utilized for an improved prediction performance for larger sample sizes.
Collapse
Affiliation(s)
- Anthony Bourached
- J. Philip Kistler Stroke Research Center, Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, USA
- UCL Queen Square Institute of Neurology, University College London, London WC1N 3BG, UK
| | - Anna K Bonkhoff
- J. Philip Kistler Stroke Research Center, Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, USA
| | - Markus D Schirmer
- J. Philip Kistler Stroke Research Center, Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, USA
| | - Robert W Regenhardt
- J. Philip Kistler Stroke Research Center, Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, USA
| | - Martin Bretzner
- J. Philip Kistler Stroke Research Center, Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, USA
- University of Lille, Inserm, CHU Lille, U1171—LilNCog (JPARC)—Lille Neurosciences & Cognition, Lille F-59000, France
| | - Sungmin Hong
- J. Philip Kistler Stroke Research Center, Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, USA
| | - Adrian V Dalca
- Computer Science and Artificial Intelligence Lab, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
- Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, Charlestown, MA 02129, USA
| | - Anne-Katrin Giese
- Department of Neurology, University Medical Center Hamburg-Eppendorf, Hamburg 20251, Germany
| | - Stefan Winzeck
- Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, Charlestown, MA 02129, USA
- Department of Computing, Imperial College London, London SW7 2RH, UK
| | - Christina Jern
- Institute of Biomedicine, Department of Laboratory Medicine, Sahlgrenska Academy, University of Gothenburg, Gothenburg 41390, Sweden
- Department of Clinical Genetics and Genomics Gothenburg, Region Västra Götaland, Sahlgrenska University Hospital, Gothenburg 41345, Sweden
| | - Arne G Lindgren
- Department of Neurology, Skåne University Hospital, Lund 22185, Sweden
| | - Jane Maguire
- Department of Clinical Sciences Lund, Neurology, Lund University, Lund 22185, Sweden
- University of Technology Sydney, Ultimo, NSW 2007, Australia
| | - Ona Wu
- Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, Charlestown, MA 02129, USA
| | - John Rhee
- Department of Neurology, Massachusetts General Hospital, Boston, MA 02139, USA
| | - Eyal Y Kimchi
- Department of Neurology, Feinberg School of Medicine, Northwestern University, Evaston, IL 60201, USA
| | - Natalia S Rost
- J. Philip Kistler Stroke Research Center, Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, USA
| |
Collapse
|