Retinal Image Processing and Computer-aided Diagnostics are included in this core research area at Visionquest.
Retinal Images
With the University of New Mexico, VisionQuest has submitted a provisional patent for the process of identifying those retinal images with various pathological structures, such as those found in diabetic retinopathy. A means for training and testing our methodology on any new data set or image format without the need for a trained ophthalmic analyst or ophthalmologist to mark each lesion has been demonstrated and is now being validated. Symposia papers describe the Amplitude Modulation-Frequency Modulation and Partial Least Squares techniques.
Pre- and post-processing (left and right image, respectively) with the lighting artifact removal software. The sharpness of the vascular network and the high contrast between lesions and retinal background makes detection more likely. The original has a fuzziness that is eliminated in the processed image. MAs are hard to distinguish in the original (left), but are easily visualized in the processed image (right).
Functional Magnetic Resonance Imaging
The objective of this grant application is to implement an adaptive, real-time functional magnetic resonance imaging (ART-FMRI) classification architecture that enables the neuroscientist to perform real-time FMRI. We differentiate our architecture from other real-time systems in that not only is the classification performed in “real-time,” but because of the one to two orders of magnitude faster training algorithm, our definition of real-time includes an ability to train and learn to recognize “new” states during an experiment. We are integrating our software into existing FMRI data handling and visualization software in order to open new applications of FMRI, including brain machine interface (BMI).
Radiology
In this R&D, we have developed a computer-aided system for reading chest radiographs according to the International Labor Organization (ILO) standards. The project introduces a methodology by which a computer-based system will screen all “normals” and refer those with suspected interstitial lung diseases or other abnormalities to the radiologist. Our system provides a reading that is based on training using exemplars that have been read by radiologists (B‑readers). Digital Radiology Image Screening System (DRISS) will screen images in seconds. The computer-assisted radiograph screening system will be implemented at several levels of automation in order to provide tailored support to the radiologist. At one level it will simply screen radiographs and at another level will provide the radiologist with estimates of profusion for each section of the lung.
