Machine learning

Goals

Machine learning introduces a framework that can help with everything from automated diagnosis to information extraction and organization. The Langlotzlab has a series of projects that work with medical images and or data, and the following are a few high level examples of what machine learning can offer…

  • diagnostic features, sometimes with human interpretability
  • dimensionality reduction of images and reports
  • generalization across modalities and body parts

For example, a standard pipeline for an “automated radiologist” might aim to use annotations of radiologist reports with a supervised algorithm to intelligently summarize the report, and possibly provide real-time decision support for radiologists:

You might also want to distinguish different components of images, as is nicely illustrated in the images below:

Projects

Diagnosis of Bone Tumors Using Radiographs