New hybrid stochastic-deterministic technique for fast registration of dermatological images

Pavlopoulos, SA

dc.contributor.author	Pavlopoulos, SA	en
dc.date.accessioned	2014-03-01T01:21:07Z
dc.date.available	2014-03-01T01:21:07Z
dc.date.issued	2004	en
dc.identifier.issn	0140-0118	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/16074
dc.subject	Dermatological applications	en
dc.subject	Image registration	en
dc.subject	Medical imaging	en
dc.subject.classification	Computer Science, Interdisciplinary Applications	en
dc.subject.classification	Engineering, Biomedical	en
dc.subject.classification	Mathematical & Computational Biology	en
dc.subject.classification	Medical Informatics	en
dc.subject.other	Algorithms	en
dc.subject.other	Dermatology	en
dc.subject.other	Image analysis	en
dc.subject.other	Optimization	en
dc.subject.other	Parameter estimation	en
dc.subject.other	Problem solving	en
dc.subject.other	Image registration	en
dc.subject.other	Medical image processing	en
dc.subject.other	Rotation parameters	en
dc.subject.other	Scaling parameters	en
dc.subject.other	Medical imaging	en
dc.subject.other	algorithm	en
dc.subject.other	article	en
dc.subject.other	cancer staging	en
dc.subject.other	controlled study	en
dc.subject.other	correlation analysis	en
dc.subject.other	diagnostic accuracy	en
dc.subject.other	diagnostic imaging	en
dc.subject.other	geometry	en
dc.subject.other	image analysis	en
dc.subject.other	image processing	en
dc.subject.other	imaging system	en
dc.subject.other	mathematical computing	en
dc.subject.other	melanoma	en
dc.subject.other	nonparametric test	en
dc.subject.other	skin examination	en
dc.subject.other	statistical analysis	en
dc.subject.other	statistical significance	en
dc.subject.other	stochastic model	en
dc.subject.other	validation process	en
dc.subject.other	Algorithms	en
dc.subject.other	Computer Simulation	en
dc.subject.other	Humans	en
dc.subject.other	Image Processing, Computer-Assisted	en
dc.subject.other	Melanoma	en
dc.subject.other	Neoplasm Staging	en
dc.subject.other	Skin Neoplasms	en
dc.subject.other	Stochastic Processes	en
dc.title	New hybrid stochastic-deterministic technique for fast registration of dermatological images	en
heal.type	journalArticle	en
heal.identifier.primary	10.1007/BF02345211	en
heal.identifier.secondary	http://dx.doi.org/10.1007/BF02345211	en
heal.language	English	en
heal.publicationDate	2004	en
heal.abstract	Digital image processing in the medical field has become very popular in recent years owing to the significant advantages it offers over conventional techniques of visual or analogue image analysis. One of the most significant aspects in medical image processing has been that of image registration, which deals with the task of registering two images taken under different conditions. Image registration is considered an important issue in the field of dermatology, as pictures of a lesion taken in different periods need to be compared and quantitatively analysed. A hybrid image registration scheme was developed and evaluated for dermatological applications. The method splits the parameter estimation problem into two, with a combination of deterministic and iterative estimation techniques. The scaling and rotation parameters are estimated using a cross-correlation of image invariant image descriptors algorithm, whereas the two translation parameters are estimated with a non-parametric similarity criterion and a hill-climbing optimisation scheme. The efficacy of the method has been validated for the registration and comparison of malignant melanoma images. Determination of rotation and scaling parameters was performed using the log-polar transformation technique, which proved to be very accurate, even when high rotation and scaling values were imposed. Deviations for the rotation parameter estimations were less than 0.5%, whereas, for the scaling factor, differences were on average less than 2.5%, with a maximum difference estimated to be 4.5%. Translation parameter estimation was performed using integer similarity measures namely the stochastic sign change, the deterministic sign change (DSC) and the window value range, the performance of which has been assessed and, in all cases, was found to be highly effective. A novel hill-climbing optimisation algorithm has been proposed and, in combination with the DSC similarity criterion, was evaluated and proved to successfully estimate translation parameters. Thus the proposed hybrid registration technique can successfully estimate problem parameters in a time-efficient manner. © IFMBE: 2004.	en
heal.publisher	PETER PEREGRINUS LTD	en
heal.journalName	Medical and Biological Engineering and Computing	en
dc.identifier.doi	10.1007/BF02345211	en
dc.identifier.isi	ISI:000225543500007	en
dc.identifier.volume	42	en
dc.identifier.issue	6	en
dc.identifier.spage	777	en
dc.identifier.epage	786	en