Hundreds Of Segments Research Articles

Strategies to Reduce the Expert Supervision Required for Deep Learning-Based Segmentation of Histopathological Images.

The emergence of computational pathology comes with a demand to extract more and more information from each tissue sample. Such information extraction often requires the segmentation of numerous histological objects (e.g., cell nuclei, glands, etc.) in histological slide images, a task for which deep learning algorithms have demonstrated their effectiveness. However, these algorithms require many training examples to be efficient and robust. For this purpose, pathologists must manually segment hundreds or even thousands of objects in histological images, i.e., a long, tedious and potentially biased task. The present paper aims to review strategies that could help provide the very large number of annotated images needed to automate the segmentation of histological images using deep learning. This review identifies and describes four different approaches: the use of immunohistochemical markers as labels, realistic data augmentation, Generative Adversarial Networks (GAN), and transfer learning. In addition, we describe alternative learning strategies that can use imperfect annotations. Adding real data with high-quality annotations to the training set is a safe way to improve the performance of a well configured deep neural network. However, the present review provides new perspectives through the use of artificially generated data and/or imperfect annotations, in addition to transfer learning opportunities.

FitEllipsoid: a fast supervised ellipsoid segmentation plugin

BackgroundThe segmentation of a 3D image is a task that can hardly be automatized in certain situations, notably when the contrast is low and/or the distance between elements is small. The existing supervised methods require a high amount of user input, e.g. delineating the domain in all planar sections.ResultsWe present FitEllipsoid, a supervised segmentation code that allows fitting ellipsoids to 3D images with a minimal amount of interactions: the user clicks on a few points on the boundary of the object on 3 orthogonal views. The quantitative geometric results of the segmentation of ellipsoids can be exported as a csv file or as a binary image. The core of the code is based on an original computational approach to fit ellipsoids to point clouds in an affine invariant manner. The plugin is validated by segmenting a large number of 3D nuclei in tumor spheroids, allowing to analyze the distribution of their shapes. User experiments show that large collections of nuclei can be segmented with a high accuracy much faster than with more traditional 2D slice by slice delineation approaches.ConclusionsWe designed a user-friendly software FitEllipsoid allowing to segment hundreds of ellipsoidal shapes in a supervised manner. It may be used directly to analyze biological samples, or to generate segmentation databases necessary to train learning algorithms. The algorithm is distributed as an open-source plugin to be used within the image analysis software Icy. We also provide a Matlab toolbox available with GitHub.

The Influence of a low level jet on the thumps generated by a wind turbine

Annoyance caused by wind turbine noise is assessed in terms of the A-weighted time average sound level. The amplitude modulation of a wind turbine, defined as the fluctuations in the noise level, increases the annoyance caused. There are two types of modulation: normal (in close proximity to the turbine) and abnormal (at a distance from the turbine). The latter is perceived as very annoying thumps. To determine the noise level of thumps, the standard simulations divide the noise source (turbine blades) into hundreds of segments. Then, taking refraction and other wave phenomena into account, the segment contributions to the sound level at the receiver are estimated. These steps are repeated for all the blade segments at each angular position of the blade. In the present study, a simple phenomenological model of noise generation and propagation is proposed. It is based on the finding that the blade tips are the major contributors to the sound power of a wind turbine. Since the abnormal modulation of wind turbine noise is heard at a long distance, refraction is taken into account for a special case of a low level jet. To apply the proposed model, two in situ noise measurements are required: within the fully insonified zone and within the partly insonified zone.

Automated home cage training of mice in a hold-still center-out reach task.

An obstacle to understanding neural mechanisms of movement is the complex, distributed nature of the mammalian motor system. Here we present a novel behavioral paradigm for high-throughput dissection of neural circuits underlying mouse forelimb control. Custom touch-sensing joysticks were used to quantify mouse forelimb trajectories with micron-millisecond spatiotemporal resolution. Joysticks were integrated into computer-controlled, rack-mountable home cages, enabling batches of mice to be trained in parallel. Closed loop behavioral analysis enabled online control of reward delivery for automated training. We used this system to show that mice can learn, with no human handling, a direction-specific hold-still center-out reach task in which a mouse first held its right forepaw still before reaching out to learned spatial targets. Stabilogram diffusion analysis of submillimeter-scale micromovements produced during the hold demonstrate that an active control process, akin to upright balance, was implemented to maintain forepaw stability. Trajectory decomposition methods, previously used in primates, were used to segment hundreds of thousands of forelimb trajectories into millions of constituent kinematic primitives. This system enables rapid dissection of neural circuits for controlling motion primitives from which forelimb sequences are built. NEW & NOTEWORTHY A novel joystick design resolves mouse forelimb kinematics with micron-millisecond precision. Home cage training is used to train mice in a hold-still center-out reach task. Analytical methods, previously used in primates, are used to decompose mouse forelimb trajectories into kinematic primitives.

Laser-guide-stars used for cophasing broad capture ranges

Segmented primary mirrors are indispensable to master the steady increase in spatial resolution. Phasing optics systems must reduce segment misalignments to guarantee the high optical quality required for astronomical science programs. Modern telescopes routinely use adaptive optics systems to compensate for the atmosphere and use laser-guide-stars to create artificial stars as bright references in the field of observation. Because multiple laser-guide-star adaptive optics are being implemented in all major observatories, we propose to use man-made stars not only for adaptive optics, but for phasing optics. We propose a method called the doublet-wavelength coherence technique (DWCT), exploiting the D lines of sodium in the mesosphere using laser guide-stars. The signal coherence properties are then used. The DWCT capture range exceeds current abilities by a factor of 100. It represents a change in paradigm by improving the phasing optics capture range from micrometric to millimetric. It thereby potentially eliminates the need of a man-made mechanical pre-phasing step. Extremely large telescopes require hundreds of segments, several of which need to be substituted on a daily basis to be recoated. The DWCT relaxes mechanical integration requirements and speeds up integration and re-integration process.

Expressive Timbre and Timing in Rhythmic Performance: Analysis of Steve Reich’s Clapping Music

Because of its minimalist construction, Steve Reich’s Clapping Music provides an ideal crucible for an examination of the role of expressive timing and expressive timbre in musical performance. Seven recordings of Clapping Music are studied. The analysis begins by dissecting the seven performances into hundreds of segments that contain either single handclaps or double handclaps. These are then studied in many ways: the consistency with which performers are able to maintain the timbre of their claps, the height percept of a handclap and its meaning for such a percussive sound, the accuracy in time with which performers can maintain a steady pulse. When clapping together, the timbre of the clap pair is dependent not only on the individual timbres but also on the timing of the claps. It is possible to measure the accuracy in time with which performers synchronize the double claps. The section-by-section timing deviations correlate significantly with a simple measure of complexity: the more complex the pattern, the greater the timing deviations.

SU‐E‐T‐08: A Convolution Model for Head Scatter Fluence in the Intensity Modulated Field

Purpose:To efficiently calculate the head scatter fluence for an arbitrary intensity‐modulated field with any source distribution using the source occlusion model.Method:The source occlusion model with focal and extra focal radiation (Jaffray et al, 1993) can be used to account for LINAC head scatter. In the model, the fluence map of any field shape at any point can be calculated via integration of the source distribution within the visible range, as confined by each segment, using the detector eye's view. A 2D integration would be required for each segment and each fluence plane point, which is time‐consuming, as an intensity‐modulated field contains typically tens to hundreds of segments. In this work, we prove that the superposition of the segmental integrations is equivalent to a simple convolution regardless of what the source distribution is. In fact, for each point, the detector eye's view of the field shape can be represented as a function with the origin defined at the point's pinhole reflection through the center of the collimator plane. We were thus able to reduce hundreds of source plane integration to one convolution. We calculated the fluence map for various 3D and IMRT beams and various extra‐focal source distributions using both the segmental integration approach and the convolution approach and compared the computation time and fluence map results of both approaches.Results:The fluence maps calculated using the convolution approach were the same as those calculated using the segmental approach, except for rounding errors (<0.1%). While it took considerably longer time to calculate all segmental integrations, the fluence map calculation using the convolution approach took only ∼1/3 of the time for typical IMRT fields with ∼100 segments.Conclusions:The convolution approach for head scatter fluence calculation is fast and accurate and can be used to enhance the online process.

The impact of internodal segmentation in biophysical nerve fiber models

Implementation of double cable models to simulate the behavior of myelinated peripheral nerve fibers requires defining a segmentation of the internode between successive nodes of Ranvier. The number of internodal segments is a model parameter that is not well agreed on, with values in the literature ranging from 1 to more than 500. Moreover, a lot of studies also lack a sensitivity study or a rationale behind the implementation used. In a model of a myelinated nerve fiber developed in our group, the segmentation scheme (i.e., the number of segments and their individual morphology) strongly influenced model outcomes such as action potential shape and velocity, stimulation threshold and absolute refractory period. In the present study these influences were investigated systematically in homogeneous neurons with different diameters. Uniformly segmented internodes were found to require several hundreds of segments (and associated computational power) to reach model outcomes differing by less than 1% from the asymptotic value. In fact, in the majority of segmentation schemes the main determinant is not the number of segments, but the length λ of the internodal segments directly adjacent to the nodes of Ranvier. If λ is larger than approximately 10 μm, model outcomes for the tested fibers are almost independent of the total number of segments. Furthermore, λ can be optimized to enable models using just three segments per internode, to reach physiologically relevant model outcomes with limited computational resources. However, to study anatomical or physiological details of the internode itself, an appropriately detailed segmentation scheme is crucial.

In-plane effects on segmented-mirror control

Extremely large optical telescopes are being designed with primary mirrors composed of hundreds of segments. The "out-of-plane" piston, tip, and tilt degrees of freedom of each segment are actively controlled using feedback from relative height measurements between neighboring segments. The "in-plane" segment translations and clocking (rotation) are not actively controlled; however, in-plane motions affect the active control problem in several important ways, and thus need to be considered. We extend earlier analyses by constructing the "full" interaction matrix that relates the height, gap, and shear motion at sensor locations to all six degrees of freedom of segment motion, and use this to consider three effects. First, in-plane segment clocking results in height discontinuities between neighboring segments that can lead to a global control system response. Second, knowledge of the in-plane motion is required both to compensate for this effect and to compensate for sensor installation errors, and thus, we next consider the estimation of in-plane motion and the associated noise propagation characteristics. In-plane motion can be accurately estimated using measurements of the gap between segments, but with one unobservable mode in which every segment clocks by an equal amount. Finally, we examine whether in-plane measurements (gap and/or shear) can be used to estimate out-of-plane segment motion; these measurements can improve the noise multiplier for the "focus-mode" of the segmented-mirror array, which involves pure dihedral angle changes between segments and is not observable with only height measurements.

On-sky multiwavelength phasing of segmented telescopes with the Zernike phase contrast sensor

Future extremely large telescopes will adopt segmented primary mirrors with several hundreds of segments. Cophasing of the segments together is essential to reach high wavefront quality. The phasing sensor must be able to maintain very high phasing accuracy during the observations, while being able to phase segments dephased by several micrometers. The Zernike phase contrast sensor has been demonstrated on-sky at the Very Large Telescope. We present the multiwavelength scheme that has been implemented to extend the capture range from ±λ/2 on the wavefront to many micrometers, demonstrating that it is successful at phasing mirrors with piston errors up to ±4.0 μm on the wavefront. We discuss the results at different levels and conclude with a phasing strategy for a future extremely large telescope.

Sedimentary facies and progradational style of a Pleistocene talus-slope succession, Northern Calcareous Alps, Austria

In mountain ranges, talus slopes are ubiquitous and typically represent the highest deposystem. The style of talus buildup from a low-dipping, immature slope to a high and steep, geomorphically mature slope, however, to date was not documented. Near Innsbruck city (Austria) a lithified talus-slope succession records progradation and downlap via talus-associated alluvial fans along its toe-of-slope; the fans linked progradation of the steep-dipping talus-slope segment over a lower-dipping substrate. The considered succession (‘Hötting Breccia’ Auct.) probably accumulated during the terminal Riss–Würm interglacial to early Würmian, and became lithified before the Last Glacial Maximum. The Hötting Breccia consists of alluvial-fan deposits which, in turn, are locally downlapped by a succession of aggrading to prograding talus slopes. Up-hill, the fossil talus slopes pinch out in onlap onto former rock cliffs. In the eastern part of outcrop, talus buildup is well-exposed along the flanks of a canyon; there, facies and depositional geometries record: (a) a basal, low-dipping alluvial-fan interval that accumulated near the toe-of-cliff, overlain and downlapped by (b) a steeper-dipping talus-slope succession. In the steep-dipping (25–35°), proximal talus-slope segment hundreds of meters in length, the talus successions consist mainly of: (i) clast-supported breccias of cohesive debris flows, intercalated with (ii) openwork breccias from grain flows and particle creep. Progradation of the steep-dipping segment of talus slopes took place via shingling of alluvial-fan depounits along the toe-of-slope. The fans linked the progradation of the steep-dipping, proximal talus-slope segment with the lower-dipping substrate ahead of the talus slope. The change from alluvial-fan deposition along the toe of initially high cliffs towards climbing onlap and progradation of talus slopes occurred when a slope segment dipping with the mean angle of residual shear of talus material had formed at the apex of the fan. Because the free cliff face supplying talus deposition diminishes with time, the progradation potential of talus slopes is inherently limited.

A Bayesian modeling of seismicity in time and space.

Basing on the Bayesian statistics, a new method to quantify seismicity is proposed. Earthquake occurrences are assumed to follow a non-stationary Poisson process of which the parameter (Poisson rate) changes both in time and in space Time-space for study is divided into hundreds of segments, and earthquakes are sorted into these segments. The Poisson rate is assumed to be constant in each segment, and the time-space variation of the rate is estimated by using Akaike's Bayesian Information Criterion (ABIC). To demonstrate the effectiveness of this method, the time-space variation in seismicity was studied for the area of the Nemuro-oki earthquake of June 17, 1973(MJMA=7.4). As a result of the analysis, seismicity in the source area is found to be rather constant for the period of 1927-1972. The precursory seismic quiescence pointed out in earlier works should be attributed to the high seismic activities in the surrounding areas due to aftershocks of the 1952, 1968, and 1969 earthquakes.

CYCLIC HETEROPHYLLY IN SYNGONIUM (ARACEAE)

Species in the genus Syngonium germinate on the ground and mature on the trunks of the trees. These vines consist of relatively unbranched shoots which grow through the forest for considerable distances both horizontally and vertically. In the three species of this study, the shape of the segment, measured as internode diam/length, falls into two distinct classes, leafy and elongate, and varies cyclicly along single shoots. Two cycles are recognized. One cycle occurs in small diam terrestrial shoots in which cycling seems to be controlled by endogenous factors, and occurs with a period of a few tens of segments. The other cycle occurs in shoots of larger diam which climb and descend from trees. In this cycle, alternation between the two forms is controlled by gain or loss of contact with trees, and shoots may remain within a phase of the cycle for hundreds of segments, as long as the substrate does not change.

Cyclic Heterophylly in Syngonium (Araceae)

Species in the genus Syngonium germinate on the ground and mature on the trunks of the trees. These vines consist of relatively unbranched shoots which grow through the forest for considerable distances both horizontally and vertically. In the three species of this study, the shape of the segment, measured as internode diam/length, falls into two distinct classes, leafy and elongate, and varies cyclicly along single shoots. Two cycles are recognized. One cycle occurs in small diam terrestrial shoots in which cycling seems to be controlled by endogenous factors, and occurs with a period of a few tens of segments. The other cycle occurs in shoots of larger diam which climb and descend from trees. In this cycle, alternation between the two forms is controlled by gain or loss of contact with trees, and shoots may remain within a phase of the cycle for hundreds of segments, as long as the substrate does not change.