Robert Brandt

Evaluation of Atlas Selection Strategies for Atlas-Based Image Segmentation with Application to Confocal Microscopy Images of Bee Brains

This paper evaluates strategies for atlas selection in atlas-based segmentation of three-dimensional biomedical images. Segmentation by intensity-based nonrigid registration to atlas images is applied to confocal microscopy images acquired from the brains of 20 bees. This paper evaluates and compares four different approaches for atlas image selection: registration to an individual atlas image (IND), registration to an average-shape atlas image (AVG), registration to the most similar image from a database of individual atlas images (SIM), and registration to all images from a database of individual atlas images with subsequent multi-classifier decision fusion (MUL). The MUL strategy is a novel application of multi-classifier techniques, which are common in pattern recognition, to atlas-based segmentation. For each atlas selection strategy, the segmentation performance of the algorithm was quantified by the similarity index (SI) between the automatic segmentation result and a manually generated gold standard. The best segmentation accuracy was achieved using the MUL paradigm, which resulted in a mean similarity index value between manual and automatic segmentation of 0.86 (AVG, 0.84; SIM, 0.82; IND, 0.81). The superiority of the MUL strategy over the other three methods is statistically significant (two-sided paired t test, P < 0.001). Both the MUL and AVG strategies performed better than the best possible SIM and IND strategies with optimal a posteriori atlas selection (mean similarity index for optimal SIM, 0.83; for optimal IND, 0.81). Our findings show that atlas selection is an important issue in atlas-based segmentation and that, in particular, multi-classifier techniques can substantially increase the segmentation accuracy.

Colour thresholds and receptor noise: behaviour and physiology compared

Photoreceptor noise sets an absolute limit for the accuracy of colour discrimination. We compared colour thresholds in the honeybee (Apis mellifera) with this limit. Bees were trained to discriminate an achromatic stimulus from monochromatic lights of various wavelengths as a function of their intensity. Signal-to-noise ratios were measured by intracellular recordings in the three spectral types of photoreceptor cells. To model thresholds we assumed that discrimination was mediated by opponent mechanisms whose performance was limited by receptor noise. Most of the behavioural thresholds were close to those predicted from receptor signal-to-noise ratios, suggesting that colour discrimination in honeybees is affected by photoreceptor noise. Some of the thresholds were lower than this theoretical limit, which indicates summation of photoreceptor cell signals.

Three-Dimensional Average-Shape Atlas of the Honeybee Brain and Its Applications

The anatomical substrates of neural nets are usually composed from reconstructions of neurons that were stained in different preparations. Realistic models of the structural relationships between neurons require a common framework. Here we present 3‐D reconstructions of single projection neurons (PN) connecting the antennal lobe (AL) with the mushroom body (MB) and lateral horn, groups of intrinsic mushroom body neurons (type 5 Kenyon cells), and a single mushroom body extrinsic neuron (PE1), aiming to compose components of the olfactory pathway in the honeybee. To do so, we constructed a digital standard atlas of the bee brain. The standard atlas was created as an average‐shape atlas of 22 neuropils, calculated from 20 individual immunostained whole‐mount bee brains. After correction for global size and positioning differences by repeatedly applying an intensity‐based nonrigid registration algorithm, a sequence of average label images was created. The results were qualitatively evaluated by generating average gray‐value images corresponding to the average label images and judging the level of detail within the labeled regions. We found that the first affine registration step in the sequence results in a blurred image because of considerable local shape differences. However, already the first nonrigid iteration in the sequence corrected for most of the shape differences among individuals, resulting in images rich in internal detail. A second iteration improved on that somewhat and was selected as the standard. Registering neurons from different preparations into the standard atlas reveals 1) that the m‐ACT neuron occupies the entire glomerulus (cortex and core) and overlaps with a local interneuron in the cortical layer; 2) that, in the MB calyces and the lateral horn of the protocerebral lobe, the axon terminals of two identified m‐ACT neurons arborize in separate but close areas of the neuropil; and 3) that MB‐intrinsic clawed Kenyon cells (type 5), with somata outside the calycal cups, project to the peduncle and lobe output system of the MB and contact (proximate) the dendritic tree of the PE1 neuron at the base of the vertical lobe. Thus the standard atlas and the procedures applied for registration serve the function of creating realistic neuroanatomical models of parts of a neural net. The Honeybee Standard Brain is accessible at www.neurobiologie.fu‐berlin.de/beebrain. J. Comp. Neurol. 492:1–19, 2005.

Differential parallel processing of olfactory information in the honeybee, Apis mellifera L.

Abstract. Two distinct neuronal pathways connect the first olfactory neuropil, the antennal lobe, with higher integration areas, such as the mushroom bodies, via antennal lobe projection neurons. Intracellular recordings were used to address the question whether neuroanatomical features affect odor-coding properties. We found that neurons in the median antennocerebral tract code odors by latency differences or specific inhibitory phases in combination with excitatory phases, have a more specific activity profile for different odors and convey the information with a delay. The neurons of the lateral antennocerebral tract code odors by spike rate differences, have a broader activity profile for different odors, and convey the information quickly. Thus, rather preliminary information about the olfactory stimulus first reaches the mushroom bodies and the lateral horn via neurons of the lateral antennocerebral tract and subsequently odor information becomes more specified by activities of neurons of the median antennocerebral tract. We conclude that this neuroanatomical feature is not related to the distinction between different odors, but rather reflects a dual coding of the same odor stimuli by two different neuronal strategies focusing different properties of the same stimulus.

Digital Atlases of the Antennal Lobe in Two Species of Tobacco Budworm Moths, the Oriental Helicoverpa assulta (Male) and the American Heliothis virescens (Male and Female)

The antennal lobe of the moth brain is the primary olfactory center processing information about pheromones and plant odors. We present here a digital atlas of the glomerular antennal lobe structures in the male of Helicoverpa assulta and the male and female of Heliothis virescens, based on synaptic antibody staining combined with confocal microscopy. The numbers of the glomeruli in the three specimens were similar, 65, 66, and 62, respectively. Whereas the male antennal lobe has a macroglomerular complex consisting of three and four units in the two species, the female lobe has two enlarged glomeruli at a corresponding position, near the entrance of the antennal nerve. Another large glomerulus, showing homology in the three specimens, is ventrally located. The small size of the heliothine moths is advantageous for confocal microscopy because the entire brain can be visualized as a single image stack. The maps are freely accessible on the internet, and the digital form of the data allows each atlas to be rotated and sectioned at any angle, providing for the identification of glomeruli in different preparations. J. Comp. Neurol. 446:123–134, 2002.

Two spatial memories for honeybee navigation

Insect navigation is thought to be based on an egocentric reference system which relates vector information derived from path integration to views of landmarks experienced en route and at the goal. Here we show that honeybees also possess an allocentric form of spatial memory which allows localization of multiple places relative to the intended goal, the hive. The egocentric route memory, which is called the specialized route memory (SRM) here, initially dominates navigation when an animal is first trained to a feeding site and then released at an unexpected site and this is why it is the only reference system detected so far in experiments with bees. However, the SRM can be replaced by an allocentric spatial memory called the general landscape memory (GLM). The GLM is directly accessible to the honeybee (and to the experimenter) if no SRM exists, for example, if bees were not trained along a route before testing. Under these conditions bees return to the hive from all directions around the hive at a speed comparable to that of an equally long flight along a trained route. The flexible use of the GLM indicates that bees may store relational information on places, connections between landmarks and the hive and/or views of landmarks from different directions and, thus, the GLM may have a graph structure, at least with respect to one goal, i.e. the hive.

Bee brains, B-splines and computational democracy: generating an average shape atlas

We describe a method to generate an average atlas from segmented 3-D images of a population of subjects. Using repeated application of an intensity-based non-rigid registration algorithm based on third-order 3-D B-splines, a sequence of average label images is created. Averaging of the non-numerical label data employs a generalization of the mode of sets of corresponding voxels, parameterized by a threshold value specifying the required level of classification confidence. The number of voxels that cannot be assigned a unique average value provides a criterion for the convergence of the iteration. For improved accuracy, efficiency, and robustness of the non-rigid registration, deformations computed during one iteration are propagated to the next iteration as initial transformation estimates. The usefulness of our method is demonstrated by applying it to generate an average atlas from segmented 3-D confocal microscopy images of 20 bee brains. We validate that the deformations found by our algorithm are meaningful by deforming the original gray-level images according to the transformations computed for the label fields.

HOW DO INSECT POLLINATORS DISCRIMINATE COLORS

ABSTRACT Basic concepts of color vision in animals and, in particular in the honeybee, are reviewed. Four models of color discrimination in honeybees are presented. Because visual systems in Hymenoptera are similar to that of the honeybee, such models can also be used to describe color discrimination in many hymenopteran pollinators. We compare predictive capacities of the models and give practical recommendations for their usage. Although models have different mathematical formulations, in most cases they give similar predictions. Examples where predictions of different models deviate are discussed.

Metric analysis of threshold spectral sensitivity in the honeybee

Behavioral spectral sensitivity curves are frequently used to characterize peripheral stages of visual processing. We test specific hypotheses about the physiology underlying honeybee spectral sensitivity by approximating published sensitivity curves with several metric models. The analysis shows that: (1) models assuming no interactions between different receptor types do not explain the behavioral data. Similarly, neither simple luminance mechanism models (sum of receptor excitations), nor models in which only the most sensitive receptor determines sensitivity fit the data. (2) The minimum number of postreceptoral mechanisms mediating discrimination is two. (3) Both mechanisms are of the chromatic type. Adding an achromatic mechanism decreases the accuracy of approximation.

The automatic pilot of honeybees.

Using scanning harmonic radar, we make visible for the first time the complete trajectories of ‘goal–vector’ flights in honeybees. We demonstrate that bees captured at an established feeding station, and released elsewhere, nevertheless embark on the previously learned vector flight that would have taken them directly home from the station, had they not been artificially displaced. Almost all of the bees maintained accurate compensation for lateral wind drift, and many completed the full length of the vector flight before starting to search for their hive. Our results showed that bees tend to disregard landscape cues during these vector flights, at least initially, and rely on the ‘optic flow’ of the ground beneath them, and their sun compass, to judge both direction and distance.