Search

We present a combined report on the results of three editions of the Cell Tracking Challenge, an ongoing initiative aimed at promoting the development and objective evaluation of cell segmentation and tracking algorithms. With 21 participating algorithms and a data repository consisting of 13 data sets from various microscopy modalities, the challenge displays today's state-of-the-art methodology in the field. We analyzed the challenge results using performance measures for segmentation and tracking that rank all participating methods. We also analyzed the performance of all of the algorithms in terms of biological measures and practical usability. Although some methods scored high in all technical aspects, none obtained fully correct solutions. We found that methods that either take prior information into account using learning strategies or analyze cells in a global spatiotemporal video context performed better than other methods under the segmentation and tracking scenarios included in the challenge.

Multi-color Fluorescent in-Situ Hybridization (M-FISH) selectively stains multiple DNA sequences using fluorescently labeled DNA probes. Proper interpretation of M-FISH images is often hampered by spectral overlap between the detected emissions of the fluorochromes. When using more than two or three fluorochromes, the appropriate combination of wide-band excitation and emission filters reduces cross-talk, but cannot completely eliminate it. A number of approaches -both hardware and software-have been proposed in the last decade to facilitate the interpretation of M-FISH images. The most used and efficient approaches use linear unmixing methods that algorithmically compute and correct for the fluorochrome contributions to each detection channel. In contrast to standard methods that require prior knowledge of the fluorochrome spectra, we present a new method, Non-Negative Matrix Factorization (NMF), that blindly estimates the spectral contributions and corrects for the overlap. Our experimental results show that its performance in terms of residual cross-talk and spot counting reliability outperforms the non-blind state-of-the-art method, the Non-Negative Least Squares (NNLS) algorithm.

We present and validate a quantitative, multidimensional image analysis protocol to assist in the early detection of lung cancer in minimal samples of bronchoalveolar lavage (BAL). To that end we stained BAL samples using Fluorescence Immunophenotyping and Interphase Cytogenetics as a Tool for the Investigation of Neoplasms (FICTION). Our method allows preliminary immunophenotypic detection of rare cancerous candidate cells, followed by accurate three-dimensional analysis of genomic integrity, to confirm or refute the initial assessment. Our results show that our automated analysis can accurately assist a human expert in the diagnostic evaluation of BAL samples.

Statistical analysis of genetic changes within cell nuclei that are far from the primary tumor would help determine whether such changes have occurred prior to tumor invasion. To determine whether the gene amplification in cells is morphologically and/or genetically related to the primary tumor requires quantitative evaluation of a large number of cell nuclei from continuous meaningful structures such as milk-ducts, tumors, etc., located relatively far from the primary tumor. To address this issue, we have designed an integrated image analysis software system for high-throughput segmentation of nuclei. Filters such as Beltrami flow-based reaction-diffusion, directional diffusion, etc., were used to pre-process the images resulting in a better segmentation. The accurate shape of the segmented nucleus was recovered using an iterative "shrink-wrap" operation. The study of two cases of ductal carcinoma in situ in breast tissue supports the biological observation regarding the existence of a preferential intraductal invasion, and therefore a common origin, between the primary tumor and the gene amplification in the cell-nuclei lining the ductal structures in the breast.

Real-time three-dimensional (3-D) reconstruction of epithelial structures in human mammary gland tissue blocks mapped with selected markers would be an extremely helpful tool for diagnosing breast cancer and planning treatment. Besides its clear clinical application, this tool could also shed a great deal of light on the molecular basis of the initiation and progression of breast cancer. We present a framework for real-time segmentation of epithelial structures in two-dimensional (2-D) images of sections of normal and neoplastic mammary gland tissue blocks. Complete 3-D rendering of the tissue can then be done by surface rendering of the structures detected in consecutive sections of the blocks. Paraffin-embedded or frozen tissue blocks are first sliced and sections are stained with hematoxylin and eosin. The sections are then imaged using conventional bright-field microscopy and their background corrected using a phantom image. We then use the fast-marching algorithm to roughly extract the contours of the different morphological structures in the images. The result is then refined with the level-set method, which converges to an accurate (subpixel) solution for the segmentation problem. Finally, our system stacks together the 2-D results obtained in order to reconstruct a 3-D representation of the entire tissue block under study. Our method is illustrated with results from the segmentation of human and mouse mammary gland tissue samples., ((c) 2004 Society of Photo-Optical Instrumentation Engineers.)