Loading...

Predicting Patient Clinical Data Using Radiomic Features

Eybposh, Mohammad Hossein | 2018

449 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 50928 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Fatemizadeh, Emadeddin
  7. Abstract:
  8. Genetic differences among patients and cancer types result in different responses to treatments and care from patients. Using personalized medicine, treatments and care can be designed with the specific needs of the patient in mind. To achieve this goal, the informative characteristics of the patient and the disease should be quantified. Quantitative Imaging or Radiomics are concerned with the characterization and quantification of the phenotypical characteristics of the tumors from medical images. Developing handcrafted features is time-consuming and requires the 3D volume of the tumor to be segmented before extracting the features. The segmentation task is considered an open problem and introduces numerous challenges to the main task. In this study, a complete Radiomic framework is developed using deep learning methods that bypasses the tumor segmentation step. In a two step framework, a bounding box is first detected around the tumor using a novel fully convolutional network. The network exploits 3D information while segmenting only 2D slices. Later, a deep multi-task learning framework based on fully convolutional networks is used to learn Radiomic features and tackle the aforementioned problems. The framework is used for patient survival prediction, which is considered a vital task in Palliative care. Also, new approaches are proposed to disentangle different factors of varioation from the raw magnetics resonance images of the brain tumors. These factors include: the degree of presence of a feature, and the spread of a feature in the image. The results showed a significant improvement in survival prediction and tumor degree classification over conventional methods that used hand-crafted features. The proposed method outperforms conventional approaches in two class survival prediction and tumor degree classification
  9. Keywords:
  10. Deep Learning ; Brain Tumor ; Magnetic Resonance Imagin (MRI) ; Radiomics ; Survival Prediction ; Fully Convolutional Networkes

 Digital Object List

 Bookmark

No TOC