Computer vision algorithms in DNA ploidy image analysis

Alexandratou, E; Sofou, A; Papasaika, H; Maragos, P; Yova, D; Kavantzas, N

dc.contributor.author	Alexandratou, E	en
dc.contributor.author	Sofou, A	en
dc.contributor.author	Papasaika, H	en
dc.contributor.author	Maragos, P	en
dc.contributor.author	Yova, D	en
dc.contributor.author	Kavantzas, N	en
dc.date.accessioned	2014-03-01T02:50:19Z
dc.date.available	2014-03-01T02:50:19Z
dc.date.issued	2006	en
dc.identifier.issn	16057422	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/35061
dc.subject	Color features	en
dc.subject	DNA ploidy	en
dc.subject	Image cytometry	en
dc.subject	Image simplification	en
dc.subject	Prostate cancer	en
dc.subject	Watershed transformation	en
dc.subject.other	Algorithms	en
dc.subject.other	Computer simulation	en
dc.subject.other	DNA	en
dc.subject.other	Image analysis	en
dc.subject.other	Microscopic examination	en
dc.subject.other	Tumors	en
dc.subject.other	DNA ploidy	en
dc.subject.other	Image cytometry	en
dc.subject.other	Image simplification	en
dc.subject.other	Prostate cancer	en
dc.subject.other	Computer vision	en
dc.title	Computer vision algorithms in DNA ploidy image analysis	en
heal.type	conferenceItem	en
heal.identifier.primary	10.1117/12.646304	en
heal.identifier.secondary	http://dx.doi.org/10.1117/12.646304	en
heal.identifier.secondary	60880O	en
heal.publicationDate	2006	en
heal.abstract	The high incidence and mortality rates of prostate cancer have stimulated research for prevention, early diagnosis and appropriate treatment. DNA ploidy status of tumour cells is an important parameter with diagnostic and prognostic significance. In the current study, DNA ploidy analysis was performed using image cytometry technique and digital image processing and analysis. Tissue samples from prostate patients were stained using the Feulgen method. Images were acquired using a digital imaging microscopy system consisting of an Olympus BX-50 microscope equipped with a color CCD camera. Segmentation of such images is not a trivial problem because of the uneven background, intensity variations within the nuclei and cell clustering. In this study specific algorithms were developed in Matlab based on the most prominent image segmentation approaches that emanate from the field of Mathematical Morphology, focusing on region-based watershed segmentation. First biomedical images were simplified under non-linear filtering (alternate sequential filters, levelings), and next image features such as gradient information and markers were extracted so as to lead the segmentation process. The extracted markers are used as seeds; watershed transformation was performed to the gradient of the filtered image. Image flooding was performed isotropically from the markers using hierarchical queues based on Beucher and Meyer methodology. The developed algorithms have successfully segmented the cell from its background and from cells clusters as well. To characterize the nuclei, we attempt to derive a set of effective color features. By analyzing more than 50 color features, we have found that a set of color features, hue, saturation-weighted hue, I1, = (R + G + B) /3, I2, = (R - B), I3 = (2G - R - B)/2, Karhunen-Loève transformation and energy operator, are effective.	en
heal.journalName	Progress in Biomedical Optics and Imaging - Proceedings of SPIE	en
dc.identifier.doi	10.1117/12.646304	en
dc.identifier.volume	6088	en