Dissertation
Predicting weight-related outcomes in healthy adolescents: clinical applications of fMRI and machine learning
Doctor of Philosophy (Ph.D.), Drexel University
Apr 2017
DOI:
https://doi.org/10.17918/etd-7301
Abstract
Obesity and obesity-related diseases have increased dramatically worldwide in recent years; however, previous studies have shown that weight loss interventions are largely ineffective in the long term. As a result, focus has shifted to determining objective predictors of weight gain, including neural correlates of such weight behaviors. Previous imaging studies have investigated the brain using univariate methods that do not enable detection of the multivariate complex patterns that may separate those prone to weight gain from those who are not. The present study used a supervised machine learning method (SVM; support vector machine) to classify adolescents (N = 135) into those who would gain weight or become weight variable over a 3-year period. Whole brain SVM analyses were performed for a) structural MRI, b) fMRI during milkshake tasting, c) fMRI during an inhibitory control go/no-go task, d) fMRI during a food image task and e) a combination of modalities. Structural scans did not significantly predict weight gain or weight variability. For functional scans in the milkshake and food image paradigms, SVMs significantly predicted weight gain using a linear mixed-effects method. Predictive accuracy increased when these two paradigms were concatenated in a single model. SVMs did not reach significance for classification of weight variability in any of the paradigms. These results support that weight gain proneness can be characterized by different neural activation to food stimuli and that these differences precede weight gain. The findings suggest that SVM may be useful for identifying neural markers of weight gain proneness.
Metrics
48 File views/ downloads
38 Record Views
Details
- Title
- Predicting weight-related outcomes in healthy adolescents
- Creators
- Samantha Winter - DU
- Contributors
- Michael R. Lowe (Advisor) - Drexel University (1970-)
- Awarding Institution
- Drexel University
- Degree Awarded
- Doctor of Philosophy (Ph.D.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Resource Type
- Dissertation
- Language
- English
- Academic Unit
- Psychological and Brain Sciences (Psychology); College of Arts and Sciences; Drexel University
- Other Identifier
- 7301; 991014632157804721