Microscopic Image Segmentation And Pathology Recognition

New methods of automated recognition of pathology patterns in biopsy and blood smear slides significantly enhance the efficiency and accuracy of pathologists’ daily work.

The microscopic study of cell morphology plays an important role in diagnosis. For a number of diseases such as infections and cancers, biopsy and blood smear diagnosis from images remains the “gold standard.” Thanks to recent advances, computer vision and machine-learning methods find a broad range of applications in biology. In our case this involves automation of the work of pathologists. A computer-aided diagnosis system performs automatic cell nuclei detection, segmentation and classification, presenting the pathologist with clear results. BitRefine group develops models capable of retrieving qualitative and quantitative information from images. We use hematoxylin-eosin and immunohistochemical staining methods in our work.

The number of pathology patterns available for automated recognition is constantly growing. Now we can work with the following tissues:

  • Breast
  • Brain
  • Blood
  • Bone marrow
  • Liver
  • Lung

Automated histopathology is still a challenging field for computer vision. However, research shows constant progress in this area. New algorithms are tested; new reference datasets are implemented for the models. The accuracy rate of the classification model differs from case to case and depends on two factors: the algorithm itself and the training dataset. BitRefine group possesses a broad range of expertise in computer vision. Together with partners from the biomedical field, we are aiming to achieve accuracy levels that will meet strict medical standards of accuracy and be accepted as part of a fully automated system.

Related Solutions

For more details, please contact us: