Machine learning classification for epilepsy lesions localization based on MRIs
Treatment of epilepsy patients usually involves antiepileptic drugs to control seizures. However, patients with drug-resistant, or refractory epilepsy go through various trials before becoming candidates for surgery to treat brain lesions that cause their seizures, and this process can take a long t...
- Autores:
-
González García, Néstor Felipe
- Tipo de recurso:
- Trabajo de grado de pregrado
- Fecha de publicación:
- 2024
- Institución:
- Universidad de los Andes
- Repositorio:
- Séneca: repositorio Uniandes
- Idioma:
- eng
- OAI Identifier:
- oai:repositorio.uniandes.edu.co:1992/73657
- Acceso en línea:
- https://hdl.handle.net/1992/73657
- Palabra clave:
- Machine learning
Epilepsy
Magnetic resonance images
Ingeniería
- Rights
- openAccess
- License
- Attribution-NonCommercial-NoDerivatives 4.0 International
| Summary: | Treatment of epilepsy patients usually involves antiepileptic drugs to control seizures. However, patients with drug-resistant, or refractory epilepsy go through various trials before becoming candidates for surgery to treat brain lesions that cause their seizures, and this process can take a long time. In this work, a machine learning approach was used to localize lesions in refractory patients by grouping lesions in two categories: hemisphere (left and right), and lobe (frontal lobe and other lobes). To train the model, a public magnetic resonance image (MRI) dataset was used and images went through a preprocessing pipeline that consisted of histogram oriented gradients, and principal component analysis. The models were tuned through cross validation techniques and then compared by their metrics to choose the best performing model for each type of image viewpoint (sagittal, coronal, and axial). |
|---|