Pattern recognition in medical imaging

点击翻译按钮获取中文摘要

Recent work in pattern recognition has demonstrated that computers can equal or even surpass image classification and pattern analysis by human experts. Modern imaging systems far exceed the human eye in spatial and spectral resolution as well as dynamic range, thus potentially allowing machine-based image pattern analysis systems to perform such tasks. The pattern analysis system we have developed, characterized and published is called WND-CHARM. A key property of WND-CHARM is that it can provide quantitative measures of image similarity. Quantitative comparisons between individual images or groups of images allow the establishment of a time-course for the progression of a physiological process represented by images. An independent verification that the continuous evolution of morphological patterns in a process are detected by the computer is its ability to place these images in order without a priori knowledge. Subsequently, the degree of change from one timepoint to another can be used to determine if the physiological process is smooth and continuous, or progresses through discrete stages. Discontinuous processes imply the presence of biological control at the transitions, and may point to key stages for medical interventions. Our published work has shown that age-related muscle degeneration (sarcopenia) in the C. elegans pharynx occurs in three discrete stages. This was the first characterization of discrete post-developmental morphological states in any organism. The observation that muscle degeneration occurs in stages implies that it may be a regulated process, and may be subject to interventions to prevent or delay these transitions. Our current work will investigate if these stages can be consistently observed in mammalian tissues, and whether interventions such as diet and the drug resveratrol affects the timing or magnitude of these morphological changes. We have also published our work on investigating the progression of osteoarthritis (OA) in the human population comprising the Baltimore Longitudinal Study of Aging (BLSA). We were able to show that WND-CHARM is able to diagnose the existence of OA in knee X-Rays with accuracies approaching that of a panel of highly trained radiologists. More recently, we have published work that WND-CHARM can predict the future onset of radiologically detectable osteoarthritis in X-Rays that were scored as radiologically clear. We were able to show that the development of moderate OA two decades in the future can be predicted with > 70% accuracy from an X-Ray scored as free of OA by a panel of three radiologists. Our current work in OA continues with the analysis of 5000 MRI scans from the Osteoarthritis Initiative. Our work with processing images of H&E-stained sections of cancer biopsies has shown that consistency of sectioning and staining is a key factor in effective image analysis of medical samples. Our initial success diagnosing and sub-classifying melanoma metastases has led us to expand this analysis to a survey of several types of cancers available as commercial tissue microarrays. Our success with this pilot survey will be followed by a more detailed study of the association between morphology/physiology and diagnosis/prognosis in prostate cancer, where a link between morphology and prognosis is already established.