MaLessforget

Chair for Computer Aided Medical Procedures & Augmented Reality
Lehrstuhl für Informatikanwendungen in der Medizin & Augmented Reality

Learning a new disease with less forgetting strategy

Supervision: Prof. Dr. Nassir Navab, Dr. Seong Tae Kim

Abstract

Recently, deep learning has great success in various applications such as image recognition, object detection, and medical applications, etc. However, in the real world deployment, the number of tasks continues to grow, or the entire tasks cannot be given at once. In other words, a new type of lesion could be defined by the medical experts and the pre-trained network needs to be re-trained to diagnose these new types of lesions. ‘Class-incremental learning’ is a research area which aims at training the learned model to add new tasks while retaining the knowledge acquired in the past tasks. It is challenging because DNNs are easy to forget previous tasks when learning new tasks (catastrophic forgetting. In real-world scenarios, it is difficult to store all training data which was used when training DNN at the previous time due to the memory and the privacy issues. In this project, we will develop a solution to this problem in medical applications by investigating effective and novel learning method.

Requirements:

  • Good understanding of statistics and machine learning methods.
  • Very good programming skills in Python & TensorFlow? / PyTorch?

Location:

  • Garching

References:

[1] Z. Zhou et al.“Fine-tuning Convolutional Neural Networks for Biomedical Image Analysis: Actively and Incrementally," CVPR 2017.

[2] S. Hou et al.“Lifelong Learning via Progressive Distillation and Retrospection," ECCV2018.

[3] F.M. Castro et al.“End-to-End Incremental Learning," ECCV2018.

[4] K. Lee et al. “Overcoming Catastrophic Forgetting with Unlabeled Data in the Wild," ICCV2019.

[5] J. Yoon et al. "Lifelong Learning for Dynamically Expandable Networks," ICLR2018

[6] W. Hu et al. "Overcoming Catastrophic Forgetting for Continual Learning via Model Adaptation," ICLR2019

If you are interested and highly-motivated to work at the intersection of machine learning and medical image analysis, please contact us via e-mail: seongtae.kim@tum.de

ProjectForm
Title: Learning a new disease with less forgetting strategy
Abstract: Recently, deep learning has great success in various applications such as image recognition, object detection, and medical applications, etc. However, in the real world deployment, the number of tasks continues to grow, or the entire tasks cannot be given at once. In other words, a new type of lesion could be defined by the medical experts and the pre-trained network needs to be re-trained to diagnose these new types of lesions. ‘Class-incremental learning’ is a research area which aims at training the learned model to add new tasks while retaining the knowledge acquired in the past tasks. It is challenging because DNNs are easy to forget previous tasks when learning new tasks (catastrophic forgetting. In real-world scenarios, it is difficult to store all training data which was used when training DNN at the previous time due to the memory and the privacy issues. In this project, we will develop a solution to this problem in medical applications by investigating effective and novel learning method.
Student:  
Director: Prof. Dr. Nassir Navab
Supervisor: Dr. Seong Tae Kim
Type: Master Thesis
Area: Machine Learning, Medical Imaging
Status: open
Start:  
Finish:  
Thesis (optional):  
Picture:  


Edit | Attach | Refresh | Diffs | More | Revision r1.2 - 02 Sep 2019 - 06:38 - SeongTaeKim