Automatic Segmentation of Unstained Living Cells in Bright-Field Microscope Images

Abstract

The Industrial Conference on Data Mining ICDM-Leipzig was the sixth event in a series of annual events which started in 2000. We are pleased to note that the topic data mining with special emphasis on real world applications has been adopted from so many researchers all over the world into their research work. We received 156 papers from 19 different countries. The main topics are data mining in medicine and marketing, web mining, mining of images and signals, theoretical aspects of data mining, and aspects of data mining that bundles a series of different data mining applications such as intrusion detection, knowledge management, manufacturing process control, time-series mining and criminal investigations. The program committee was working hard in order to select the best papers. The acceptance rate was 30%. All these selected papers are published in this proceeding volume as long papers up to 15 pages. Besides that we installed a forum where work in progress has been presented. These papers are collected in a special poster proceeding volume and show once more the potentials and interesting developments for data mining for different applications. Three new workshops have been established in connection with ICDM: 1. Mass Data Analysis on Images and Signals, MDA 2006, 2. Data Mining for Life Sciences, DMLS 2006, and 3. Data Mining in Marketing, DMM 2006. These workshops are developing new topics for data mining under the aspect of the special application. We are pleased to see how many interesting developments are going on under these topics. We would like to express our appreciation to the reviewers for their precise and highly professional works. We appreciate the help and understanding of the editorial staff at Springer and in particular Alfred Hofmann, who supported the publication of these proceedings in the LNAI series. We wish to thank all speakers, participants, and industrial exhibitors who contributed to the success of the conference. We are looking forward to welcoming you to ICDM 2007 (www.data-mining-forum.de) and to the new work you will present there.

Reconstruction of Cellular Biological Structures from Optical Microscopy Data

“…This approach is fragile because small islands of noisy pixels can cause over-segmentations. Tscherepanow et al [25] presented an active contour approach with a snakes formulation wherein overlapping cells are segmented as dictated by a 2D elliptical model. Jones et al [4] presented a novel method that first defines a metric in the image plane for calculating distances from seed regions.…”

“…All of the approaches discussed above rely on using information derived either from cellular shape models [25] or internuclear gradients [4] or neck shape cues [24]. It is interesting to note that no single approach elegantly incorporates all the cues into the segmentation process.…”

“…Identifying each nucleus separately in a biologically consistent fashion is non-trivial. While some histological stains provide viable clues in the form of sharp color-space gradients at the boundaries [4], others exhibit a narrow neck at the site of overlap between two nuclei [25]. Please refer, for example, to Figure 1(b) wherein a pair of nuclei are shown with color gradients and a narrow neck indicated by the white arrows.…”

Automatic Cell Detection in Bright-Field Microscope Images Using SIFT, Random Forests, and Hierarchical Clustering

“…Unstained cell recognition in bright-field images is a challenging problem [20], [21], [36], [38], [42]. Cells exhibit a great diversity in shape and size.…”

Automated Analysis and Classification of Infected Macrophages Using Bright-Field Amplitude Contrast Data

“…Several previous object recognition techniques have attempted to detect and count sf9 cells in bright-field images using active contour models. 4,5 Active contour models do not work satisfactorily in the presence of strong texture and pixel intensity variation, which is a characteristic of the amplitude contrast images. The broader literature review shows that not much work has been done on automatic analysis of bright-field cell data.…”