Deep Learning for Pathology Detection in Chest X-Rays
Principal Investigator
Name
Ivo Matteo Baltruschat
Degrees
M.Sc.
Institution
University Medical Center Hamburg-Eppendorf
Position Title
Research assistant
Email
About this CDAS Project
Study
PLCO
(Learn more about this study)
Project ID
PLCO-275
Initial CDAS Request Approval
May 24, 2017
Title
Deep Learning for Pathology Detection in Chest X-Rays
Summary
Nowadays, X-ray imaging is the most commonly employed procedure in a radiology department. IN particular for ICU patients, chest radiographs are frequently utilised for the detection and differentiation of pathologies such as pleural effusion, inflammatory infiltrates, pulmonary congestion, cardiac enlargement or pneumothorax and the detection of forge in bodies. Even for experienced radiologists, the identification and discrimination of these diseases is a non-trivial tasks, and decision support systems provide promising option for an optimised radiology workflow and an improved quality of care.
In recent years, Deep Learning techniques like Convolutional Neural Networks (CNN), received considerable attention in many scientific domains from computer vision to speech processing. CNN-based approaches hold leading positions in scientific contest such as the "ImageNet Large Scale Visual Recognition Competition", while the applicability of the techniques in the medical domain is subject of ongoing research.
The aim of this project is to investigate and develop techniques for the analysis of chest radiographs with a special focus on Deep Learning. Our objective is to apply deep learning methods to high-resolution medical x-ray images. In the long-term, we are pursuing the goal to employ deep learning methods in order to find patterns between medical reports and medical x-ray images that will help to push the state-of-the-art in computer-aided diagnosis for chest x-ray images.
In recent years, Deep Learning techniques like Convolutional Neural Networks (CNN), received considerable attention in many scientific domains from computer vision to speech processing. CNN-based approaches hold leading positions in scientific contest such as the "ImageNet Large Scale Visual Recognition Competition", while the applicability of the techniques in the medical domain is subject of ongoing research.
The aim of this project is to investigate and develop techniques for the analysis of chest radiographs with a special focus on Deep Learning. Our objective is to apply deep learning methods to high-resolution medical x-ray images. In the long-term, we are pursuing the goal to employ deep learning methods in order to find patterns between medical reports and medical x-ray images that will help to push the state-of-the-art in computer-aided diagnosis for chest x-ray images.
Aims
Using deep learning to solve the following tasks in a cascaded approach (get as far as possible until end of thesis work):
- Normal or Abnormal classification in chest X-ray scans
- Benchmark against existing detection algorithms
- Pathology classification (Infiltrate, effusion, stasis, tumor, pneumothorax, heart size, foreign objects)
- Integrate other information (CRP value, medical history, …) than pixel values
- Evaluate self-trained vs. pre-trained deep learning networks to solve above mentioned tasks
Collaborators
Prof. Dr.-Ing. Tobias Knopp, Institute of Biomedical Imaging