Computer vision can be a powerful tool for the analysis of medical data. The goal of image analysis tools is not to replace the medical doctor, but to assist him/her in the following activities:
enhancing the quality of image data: one common problem of medical imaging data is noise, often due to the technology, or due to the micro-movements of the patient during the capturing procedure. Computer vision can help in reducing these factors and to further clarify the visualization of imaging data through enhanced representations
simplifying the visualization of complex 3D structures: image data in medical images are often volumes of data, created through multiple scans (think of MRI for example). These volumes are dense “cubes” of information that are usually very difficult to visualize and analyze. 3D volume segmentation can greatly help in isolating the important parts of the volume to clean the data from the uninformative clutter.
measuring the extent of revealing factors in diseases: machine learning and deep learning algorithms can learn the discriminative visual markers of a particular disease, and can help the medical doctors to quantify these markers for more accurate diagnoses.
Common application fields of computer vision in medical imaging are:
Dermatology: analysis of dermoscopy image data, Optical Coherence Tomography (OCT) scans, confocal microscopy images
Magnetic Resonance Imaging (MRI) for brain or hearth diseases
Pikkart’s R&D team has years of experience in medical image analysis, with a focus on 3D volumes, 3D structure segmentation and image denoising.
“Dynamic Optical Coherence Tomography in Dermatology”, Ulrich et al., Dermatology, 2016
“Skin Surface Reconstruction and 3D Vessels Segmentation in Speckle Variance Optical Coherence Tomography”, Manfredi et al., VISAPP, 2016