Volume 8 Supplement 1
A visual model approach to extract regions of interest in microscopical images of basal cell carcinoma
© Gutiérrez and Romero; licensee BioMed Central Ltd. 2013
Published: 30 September 2013
The virtual microscopy is a discipline that emulates the interaction between an expert with a microscopical sample upon a high resolution digital slide . This type of technology is used for medical training and medical education, but so far it has been exclusively used in research environments because of the large computational requirements . For instance, after digitizing 1cm2 of a physical slide at a level of ×20 magnification, the resulting virtual slide amounts to a 4GB , that must be processed, transmitted, explored and analyzed in real time. This picture can be even worsen since a typical laboratory takes hundreds of slides each week .
Overall, a typical pathologist does not explore the entire slide, but instead she/he focus her/his analysis on a few number of visual fields or regions of interest (RoI ). In consequence, recognition of RoI in microscopical images may be a potential source of knowledge in many diagnostic tasks. Such RoIs would introduce new learning paradigms that would be used in medical education, medical training and diagnosis assistance. In addition, a precise determination of such regions can highly reduce the computational and transmission charge of informative regions from a sample. However, automatic recognition of such regions is really a challenging task because of the inherent randomness of tissue’s cutting, color tissue properties and tissue orientation.
In spite of these difficulties, the pathologist efficiently recognizes regions of interest in several domains by fusing image and task dependent information into a unique framework. This paper proposes a novel automatic approach to recognize RoIs by emulating the processing of the human visual system (HVS), not only modeling the preattentional process but also integrating it with high level processes. Hence, this paper extends our previous work  by including structural information about the relationships between several objects and texture recognition as higher cortex functions. These processes are necessary to minimally perceive the core of a scene, just as it is carried out within the pathologist memory , and therefore, to identify relevant regions for diagnosis.
Material and methods
The model was tested with a total of 115 histological microscopical fields of view of different types of basal cell carcinoma, sampled from 25 randomly chosen patients. Each biopsy was formalin-fixed and stained with Hematoxylin-Eosin dyes. Microscopical fields were digitized with a Nikon eclipse E600 system, through a coupled Nikon DXM1200 camera, and stored in JPEG format at a 1280 × 1024 resolution using microscope magnifications of ×4, ×10, ×20 and ×40. An expert pathologist, with at least five years of experience, selected the digitized fields of view and manually segmented relevant regions. We use 20 images to extract textons of size 32 × 32 pixels from RoIs and background for the object recognition task, and 95 images to test the entire algorithm.
Then, this intermediate map is used to feed a module that efficiently looks for basal cell carcinoma by comparing the pattern composition with a data base of carcinoma regions, starting with the highest attentional region (inferotemporal gyrus function). Once the algorithm sets up the first carcinoma region within the image, the other carcninoma regions are defined as the most similar regions using an Euclidean metrics of the different basic features: color, entropy and image intensity.
Structure search and recognition
During a searching task of an undetermined number of targets, the pathologist determine whether she/he is familiar with the scene, using simple features as color and texture information, possibly recognizing some relevant objects and their relations [6, 8]. If this is the case, she/he efficiently looks for the targets, following the saliency of the scene and then using the target contextual relationships [8, 9]. Likewise, our proposal follows the most saliency levels of the map obtained in the local recognition step , that is say the third stage, when the system has determined a target. We extract then some textons from RoIs previously marked (object recognition) and some from the background (basic background features extraction) to discard irrelevant structures. A classical kNN algorithm was used, using the low level features captured before as input . In case of finding a target, its relations with the other structures in the image are determined, comparing how different its internal features are. Formally, the distance between features (df ) within the target structure (st ) and the i-structure (si ) is defined as the Euclidean mean distance of the intensity features (I), orientation (O), color (C) and entropy(H) inside each structure, formally, 0 identifying similar structures as part of the searched targets using a simple threshold. Note that this algorithm addresses the intensive computations to a few structures, until it reaches the first target.
Results and discussion
Mean (variance) [%]
Gutiérrez et al. (2011)
This paper presented a novel methodology to find RoI based on the human visual system. This differs from our previous approach by the inclusion of a stage of region recognition and evaluation of inter-region similarity. These characteristics let us improve the RoI extraction since the selection criteria are modified by a knowledge database.
List of abbreviations
Human Visual System
Region of Interest
This work was partially funded by de project “Sistema para la recuperación de imágenes médicas utilizando indexación multimodal” Grant: 110152128767 from “CONVOCATORIA COLCIENCIAS 510 de 2010”
- Catalyürek U, Beynon MD, Chang C, Kurc T, Sussman A, Saltz J: The Virtual Microscope. IEEE Trans. Inf. Technol. Biomed. 2003, 7 (4): 230-248. 10.1109/TITB.2004.823952.View ArticlePubMed
- Tsuchihashi Y, Takamatsu T, Hashimoto Y, Takashima T, Nakano K, Fujita S: Use of virtual slide system for quick frozen intra-operative telepathology diagnosis in Kyoto, Japan. Diagnostic Pathology. 2008, 3 (Suppl 1): S6-10.1186/1746-1596-3-S1-S6.PubMed CentralView ArticlePubMed
- Iregui M, Gómez F, Romero E: Strategies for efficient virtual microscopy in pathological samples using JPEG2000. Micron. 2007, 38 (7): 700-713. 10.1016/j.micron.2007.04.008.View ArticlePubMed
- Doyle S, Monaco J, Madabhushi A, Lindholm S, Tomaszewski J: Evaluation of effects of JPEG2000 compression on a computer-aided detection system for prostate cancer on digitized histopathology. IEEE International Symposium on Biomedical Imaging: From Nano to Macro. 2010, 1313-1316.View Article
- Gutiérrez R, Gómez F, Roa-Peña L, Romero E: A supervised visual model for finding regions of interest in basal cell carcinoma images. Diagnostic Pathology. 2011, 6 (26): [http://www.diagnosticpathology.org/content/6/1/26]
- Wolfe JM: What do you know about what you saw?. Current Biology. 1998, 8 (9): 303-304. 10.1016/S0960-9822(98)70192-7.View Article
- Itti L, Koch C, Niebur E: A model of saliency-based visual attention for rapid scene analysis. IEEE Transactions on patern analysis and machine intelligence. 1998, 20: 1254-1259. 10.1109/34.730558.View Article
- Wolfe JM, Horowitz TS, Michod KO: Is visual attention required for robust picture memory?. Vision Research. 2007, 47: 955-964. 10.1016/j.visres.2006.11.025.PubMed CentralView ArticlePubMed
- Kunar MA, Flusberg S, Wolfe JM: The role of memory and restricted context in repeated visual search. Perception & Psychophysics. 2008, 70 (2): 314-328. 10.3758/PP.70.2.314.View Article
- Cover T, Hart P: Nearest neighbor pattern classification. Information theory, IEEE Transactions. 1967, 13: 21-27.View Article
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.