Model selection for applied machine learning research

A clinician relies on multimodal data: diagnostic scans, modifiable risk factors, non-modifiable risk factors, and patient medical history, to treat a patient for a given neurological condition. Naturally, machine learning models designed to predict patient outcomes must integrate multi-modal information to be comprehensive and reliable in a clinical setting. Furthermore, for a given dataset the performance and generalizability of various machine learning models differs based on their robustness to noise and richness of input features.

This emphasizes model selection, an important design choice in applied machine learning. The selection of features that most heavily influenced the model and the predictive performance of the model are the two main factors that determine the best model. Through empirical feature selection, we are able to decrease the complexity of our machine learning models while maintaining a high degree of predictive accuracy. These top-ranking features are then validated with a domain expert for clinical relevance as a means of enhancing model interpretability.

Helen L. Carlson, Brandon T. Craig, Alicia Hilderley, Jacquie Hodge, Deepthi Rajashekar, Pauline Mouches, Nils D. Forkert, Adam Kirton:
Structural and functional connectivity of motor circuits after perinatal stroke: A machine learning study.
In: NeuroImage: Clinical 28, Pg: 102508, 2020
DOI:10.1016/j.nicl.2020.102508

Sascha Gill, Pauline Mouches, Sophie Hu, Deepthi Rajashekar, Frank P. MacMaster, Eric E. Smith, Nils D. Forkert, Zahinoor Ismail:
Using machine learning to predict dementia from neuropsychiatric symptom and neuroimaging data.
In: Journal of Alzheimer's Disease, 75(1), 277-288, 2020.
DOI: 10.3233/JAD-191169

Lucas Lo Vercio, Kimberly Amador, Jordan J. Bannister, Sebastian Crites, Alejandro Gutierrez, M. Ethan MacDonald, Jasmine Moore, Pauline Mouches, Deepthi Rajashekar, Serena Schimert, Nagesh Subbanna, Anup Tuladhar, Nanjia Wang, Matthias Wilms, Anthony Winder, Nils D. Forkert:
Supervised machine learning tools: a tutorial for clinicians.
In: Journal of Neural Engineering, 2020
DOI: 10.1088/1741-2552/abbff2

Sarah J. MacEachern, Nils D. Forkert:
Machine learning for precision medicine
Genome, 2020 [Epub ahead of print]

Rani G. Sah, Samaneh Nobakht, Deepthi Rajashekar, Pauline Mouches, Nils D. Forkert, Amith Sitaram, Adrian Tsang, Michael D. Hill, Andrew M. Demchuk, Christopher D. d’Esterre, Philip A. Barber:
Temporal evolution and spatial distribution of quantitative T2 MRI following acute ischemia reperfusion injury.
In: International Journal of Stroke, 15(5), 495-506, 2020.
DOI: 10.1177/1747493019895673

Helen, L. Carlson, Brandon T. Craig, Jacquie Hodge, Deepthi Rajashekar, Pauline Mouches, Nils D. Forkert, Adam Kirton:
Neuroimaging can predict personalized motor function after perinatal stroke: A machine learning study.
26th Annual Meeting of the Organization for Human Brain Mapping, Montreal, Canada, 2020.

Trevor A. Seeger, Jason Tabor, Stacy Sick, Kathryn J. Schneider, Craig Jenne, Parker La, Aron S. Talai, Deepthi Rajashekar, Pauline Mouches, Nils D. Forkert, Carolyn Emery, Chantel T. Debert:
The association of saliva cytokines and pediatric sport-related concussion outcomes.
In: Journal of Head Trauma Rehabilitation, 35(5), 354-362, 2020.
DOI:10.1097/HTR.0000000000000605

Sarah J. MacEachern, Deepthi Rajashekar, Pauline Mouches, Nathan C. Rowe, Emily McKenna, Kristen W. Yeom, Nils D. Forkert:
Image-based classification of children with autism spectrum disorder.
In: 34th International Computer Assisted Radiology and Surgery (CARS) Congress, Munich, Germany, 2020.

M. Ethan MacDonald, Rebecca J. Williams, Deepthi Rajashekar, Randall B. Stafford, Alexadru Hanganu, Hongfu Sun, Avery J.L. Berman, Cheryl M. McCreary, Richard Frayne, Nils D. Forkert, G. Bruce Pike:
Age-related differences in cerebral blood flow and cortical thickness with an application to age prediction
In: Neurobiology of Aging, 95, 131-142, 2020.
DOI: 10.1016/j.neurobiolaging.2020.06.019

M. Ethan MacDonald, Deepthi Rajashekar, Rebecca J. Williams, Hongfu Sun, Cheryl McCreary, Richard Frayne, Nils D. Forkert, G. Bruce Pike:
Machine learning methods for age prediction using cortical thickness and cerebral blood flow.
25th Annual Meeting of the Organization for Human Brain Mapping, Rome, Italy, 2019.

Pauline Mouches
Deepthi Rajashekar
Kimberly Amador
Jasmine Moore

Model selection for applied machine learning research

Publications

Team members

Pauline Mouches

Deepthi Rajashekar

Kimberly Amador

Jasmine Moore