Staining the mRNA of a gene via in situ hybridization (ISH) during the development of a D. melanogaster embryo delivers the detailed spatio-temporal pattern of expression of the gene. Many biological problems such as the detection of co-expressed genes, co-regulated genes, and transcription factor binding motifs rely heavily on the analyses of these image patterns. The increasing availability of ISH image data motivates the development of automated computational approaches to the analysis of gene expression patterns.

Landmark correspondences can be used for various tasks in image processing such as image alignment, reconstruction of panoramic photographs, object recognition and simultaneous localization and mapping for mobile robots. The computer vision community knows several techniques for extracting and pairwise associating such landmarks using distinctive invariant local image features. Two very successful methods are the Scale Invariant Feature Transform (SIFT)1 and Multi-Scale Oriented Patches (MOPS).2
We implemented these methods in the Java programming language3 for seamless use in ImageJ.4 We use it for fully automatic registration of gigantic serial section Transmission Electron Microscopy (TEM) mosaics. Using automatically detected landmark correspondences, the registration of large image mosaics simplifies to globally minimizing the displacement of corresponding points.
We present here an introduction to automatic landmark correspondence detection and demonstrate our implementation for ImageJ. We demonstrate the application of the plug-in on diverse image data.

Full reconstruction of neuron morphology is of fundamental interest for the analysis and understanding of neuron function. We have developed a novel method capable of tracing neurons in three-dimensional microscopy data automatically. In contrast to template-based methods, the proposed approach makes no assumptions on the shape or appearance of neuron’s body. Instead, an efficient seeding approach is applied to find significant pixels almost certainly within complex neuronal structures and the tracing problem is solved by computing an graph tree structure connecting these seeds. In addition, an automated neuron comparison method is introduced for performance evaluation and structure analysis. The proposed algorithm is computationally efficient. Experiments on different types of data show promising results.

Mapping the connectivity of neurons in the brain (i.e., connectomics) is a challenging problem due to both the number of connections in even the smallest organisms and the nanometer resolution required to resolve them. Because of this, previous connectomes contain only hundreds of neurons, such as in the C.elegans connectome. Recent technological advances will unlock the mysteries of increasingly large connectomes (or partial connectomes). However, the value of these maps is limited by our ability to reason with this data and understand any underlying motifs. To aid connectome analysis, we introduce algorithms to cluster similarly-shaped neurons, where 3D neuronal shapes are represented as skeletons. In particular, we propose a novel location-sensitive clustering algorithm. We show clustering results on neurons reconstructed from the Drosophila medulla that show high-accuracy.

Gene expression patterns obtained by in situ mRNA hybridization provide important information about different genes during Drosophila embryogenesis. So far, annotations of these images are done by manually assigning a subset of anatomy ontology terms to an image. This time-consuming process depends heavily on the consistency of experts.

The development of high-resolution microscopy makes possible the high-throughput screening of cellular information, such as gene expression at single cell resolution. One of the critical enabling techniques yet to be developed is the automatic recognition or annotation of specific cells in a 3D image stack. In this paper, we present a novel graph-based algorithm, ARC, that determines cell identities in a 3D confocal image of C. elegans based on their highly stereotyped arrangement. This is an essential step in our work on gene expression analysis of C. elegans at the resolution of single cells. Our ARC method integrates both the absolute and relative spatial locations of cells in a C. elegans body. It uses a marker-guided, spatially-constrained, two-stage bipartite matching to find the optimal match between cells in a subject image and cells in 15 template images that have been manually annotated and vetted. We applied ARC to the recognition of cells in 3D confocal images of the first larval stage (L1) of C. elegans hermaphrodites, and achieved an average accuracy of 94.91%.

Digital reconstruction of 3D neuron structures is an important step toward reverse engineering the wiring and functions of a brain. However, despite a number of existing studies, this task is still challenging, especially when a 3D microscopic image has low single-to-noise ratio and discontinued segments of neurite patterns.

Automatic segmentation of nuclei in 3D microscopy images is essential for many biological studies including high throughput analysis of gene expression level, morphology, and phenotypes in single cell level. The complexity and variability of the microscopy images present many difficulties to the traditional image segmentation methods. In this paper, we present a new method based on 3D watershed algorithm to segment such images. By using both the intensity information of the image and the geometry information of the appropriately detected foreground mask, our method is robust to intensity fluctuation within nuclei and at the same time sensitive to the intensity and geometrical cues between nuclei. Besides, the method can automatically correct potential segmentation errors by using several post-processing steps. We tested this algorithm on the 3D confocal images of C.elegans, an organism that has been widely used in biological studies. Our results show that the algorithm can segment nuclei in high accuracy despite the non-uniform background, tightly clustered nuclei with different sizes and shapes, fluctuated intensities, and hollow-shaped staining patterns in the images.

We present an algorithm for automatic segmentation of the human pelvic bones from CT datasets that is based on the application of a statistical shape model. The proposed method is divided into three steps: 1) The averaged shape of the pelvis model is initially placed within the CT data using the Generalized Hough Transform, 2) the statistical shape model is then adapted to the image data by a transformation and variation of its shape modes, and 3) a final free-form deformation step based on optimal graph searching is applied to overcome the restrictive character of the statistical shape representation. We thoroughly evaluated the method on 50 manually segmented CT datasets by performing a leave-one-out study. The Generalized Hough Transform proved to be a reliable method for an automatic initial placement of the shape model within the CT data. Compared to the manual gold standard segmentations, our automatic segmentation approach produced an average surface distance of 1.2 ± 0.3mm after the adaptation of the statistical shape model, which could be reduced to 0.7±0.3mm using a final free-form deformation step. Together with an average segmentation time of less than 5 minutes, the results of our study indicate that our method meets the requirements of clinical routine.

In this paper, we present an automatic method for estimating the trajectories of Escherichia coli bacteria from in vivo phase-contrast microscopy videos. To address the low-contrast boundaries in cellular images, an adaptive kernel-based technique is applied to detect cells in sequence of frames. Then a novel matching gain measure is introduced to cope with the challenges such as dramatic changes of cells’ appearance and serious overlapping and occlusion. For multiple cell tracking, an optimal matching strategy is proposed to improve the handling of cell collision and broken trajectories. The results of successful tracking of Escherichia coli from various phase-contrast sequences are reported and compared with manually-determined trajectories, as well as those obtained from existing tracking methods. The stability of the algorithm with different parameter values is also analyzed and discussed.