University of Nottingham
  • Print
  

Image analysis for diagnosis

Estimating the gestational age of preterm infants using machine learning and computer vision

The detail in biomedical images provides a rich source of information about the body, and new imaging techniques are revealing increasing levels of detail. The University of Nottingham has the largest database of radiological images in the UK which will become available to researchers all over the world. 

We are developing new image analysis techniques to extract clinically relevant information from these images to help with increasingly accurate clinical diagnosis and decision support, using our strong clinical links to validate our analysis and ensure they are clinically applicable.

Technical Capabilities

With research capabilities covering many disciplines, we work in the fields of 

  • Machine Learning
  • Image Processing
  • Computer Vision
  • Artificial Intelligence
  • Motion Analysis
  • Medical Image Segmentation
  • Shape and Object Data Analysis

Clinical applications

Motion analysis in radiology imaging

  • We can track the shape and motion of bones in articulated joints from CT images, reconstruct these in 3D, and detect abnormalities, which has been applied to wrist pathology. (Xin Chen)
  • Motion analysis has also been applied to respiratory motion correction in MRI/PET images. (Xin Chen)
 

Image segmentation and machine learning

 

  • We have developed facial recognition system and eye tracking software which is being used for ADHD and autism diagnosis (Michel Valstar)
  • We have developed a simple to use mobile phone app to better estimate gestational age, which will be used to support premature babies in low income countries (Michel Valstar, Mercedes Torres Torres)
  • We have developed an automatic nerve fibre segmentation and quantification tool which enables early detection of peripheral neuropathy for diabetic patients using corneal confocal microscopic images. (Xin Chen)
  • We have developed automatic segmentation of fat, glandular tissue, pectoral muscle and nipple in breast mammograms, which further lead to the automatic estimation of 3-D volumetric breast density based on a single 2-D digital mammogram. (Xin Chen)
 

Using shape and object data analysis we have worked on:

 

  • Diagnosing idiopathic lung fibrosis from lung CT scans Expert: Chris Brignell
  • Identifying the composition of pharmaceutical tablets from Raman spectroscopy data . Expert: Chris Brignell
  • Biomarker discovery and evaluation. Expert: John Robertson
  • Early detection of breast cancer and colon cancer
 
 

Related Centres, Institutes and Groups

Computer Vision Laboratory Medical Imaging Unit MindTech Mixed Reality Laboratory Nottingham BRC Nottingham Molecular Pathology Node Radiological Sciences Research Group Shape and Object Data Analysis Sir Peter Mansfield Imaging Centre