Show simple item record

dc.contributor.advisorElfar Adalsteinsson and Jayashree Kalpathy-Cramer.en_US
dc.contributor.authorKo, Sean,M. Eng.Massachusetts Institute of Technology.en_US
dc.contributor.otherMassachusetts Institute of Technology. Department of Electrical Engineering and Computer Science.en_US
dc.date.accessioned2020-09-15T21:56:50Z
dc.date.available2020-09-15T21:56:50Z
dc.date.copyright2020en_US
dc.date.issued2020en_US
dc.identifier.urihttps://hdl.handle.net/1721.1/127418
dc.descriptionThesis: M. Eng., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, May, 2020en_US
dc.descriptionCataloged from the official PDF of thesis.en_US
dc.descriptionIncludes bibliographical references (pages 63-73).en_US
dc.description.abstractWhite matter lesion detection in the brain is crucial in diagnosing cerebral adrenoleukodystrophy early in order for life-saving therapies to be as effective as possible. To assist doctors in this early diagnosis, I developed an end-to-end deep learning pipeline to automatically segment these brain lesions in MRI scans. These segmentations will be able to help automatically compute lesion volume and risk scores to help recommend best course of treatment. This pipeline will be optimized by tuning different preprocessing and training parameters to this task. The former half of the pipeline is comprised of all the preprocessing steps performed on the raw data to prepare it as model inputs and reduce training hindrances in the images themselves. These include data conversion, RAI orientation, isotropic resampling, N4 bias correction, coregistration, skull stripping, and normalization. The latter portion of the pipeline focuses on optimizing the training of a 3D UNet by tuning different training parameters, such as patch size, loss function, normalization layers, transfer learning, and additional input channels. The scope of the findings presented here is not limited to this specific task and may be extended to other segmentation tasks as well.en_US
dc.description.statementofresponsibilityby Sean Ko.en_US
dc.format.extent73 pagesen_US
dc.language.isoengen_US
dc.publisherMassachusetts Institute of Technologyen_US
dc.rightsMIT theses may be protected by copyright. Please reuse MIT thesis content according to the MIT Libraries Permissions Policy, which is available through the URL provided.en_US
dc.rights.urihttp://dspace.mit.edu/handle/1721.1/7582en_US
dc.subjectElectrical Engineering and Computer Science.en_US
dc.titleOptimizing a deep learning approach for automatic segmentations for white matter lesionsen_US
dc.typeThesisen_US
dc.description.degreeM. Eng.en_US
dc.contributor.departmentMassachusetts Institute of Technology. Department of Electrical Engineering and Computer Scienceen_US
dc.identifier.oclc1192561553en_US
dc.description.collectionM.Eng. Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Scienceen_US
dspace.imported2020-09-15T21:56:49Zen_US
mit.thesis.degreeMasteren_US
mit.thesis.departmentEECSen_US


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record