MRI Segmentation

Machine learning and computer vision techniques allow for performance of automatic segmentation of MRI and extraction of volumetric and shape data for specific anatomic regions.

Segmentation in MRI is the first part of quantitative image analysis of different structures. Now, in clinical practice as well as in research labs, segmentation is done manually, which is very time-consuming. Recent research shows the ability of modern models such as artificial neural networks to successfully deal with such high-level tasks as biological image processing. BitRefine group performs continuous study of artificial neural network architectures and parameter variations to increase segmentation accuracy and add more regions for automatic detection.

Now our efforts are mainly related to brain MRI:

  • Brain tumor segmentation
  • Midbrain structures segmentation

We’re carrying out research on MRI of different structures such as kidney, prostate, and others.

Although our approach shows great potential in MRI segmentation, the recognition model is not universally applicable. Different organs or tumors may need parametric tuning, model retraining or even choosing a different architecture or method. Thus, BitRefine group offers R&D services, adapting our solutions for individual tasks. Developing a model for a particular object may also require a massive number of annotated images for training. We offer adaptive solutions that can continue training after deployment, continually increasing recognition accuracy.

Related Solutions

For more details, please contact us: