A multi-resolution textural approach to diagnostic neuropathology reporting

摘要: We present a computer aided diagnostic workflow focusing on two branch points in neuropathology (intraoperative consultation and p53 status tumor biopsy specimens) by means of texture analysis via discrete wavelet frames decomposition. For intraoperative consultation, our methodology is capable classifying glioblastoma versus metastatic cancer extracting textural features from the non-nuclei region cytologic preparations based imaging characteristics glial processes, which appear as anisotropic thin linear structures. metastasis, these are homogeneous appearance, thus suitable extractable distinguish tissue types. Experiments 53 images (29 glioblastomas 24 metastases) resulted average accuracy high 89.7 % for glioblastoma, 87.5 % metastasis 88.7 % overall. interpretation, we detect classify staining intensity into strong, moderate, weak negative sub-classes. achieved this developing novel adaptive thresholding detection, two-step rule weighted color classification positively negatively stained nuclei, followed to nuclei moderate Our detection method able correctly locate four types cells, at 85 % precision 88 % sensitivity rate. These methods other hand recorded 81 % positive 60 % further cells three groups, comparable with neuropathologists' markings.