Automation of Caenorhabditis elegans lifespan assay using a simplified domain synthetic image-based neural network training strategy.

Performing lifespan assays with Caenorhabditis elegans (C. elegans) nematodes manually is a time consuming and laborious task. Therefore, automation is necessary to increase productivity. In this paper, we propose a method to automate the counting of live C. elegans using deep learning. The survival curves of the experiment are obtained using a sequence formed by an image taken on each day of the assay. Solving this problem would require a very large labeled dataset; thus, to facilitate its generation, we propose a simplified image-based strategy. This simplification consists of transforming the real images of the nematodes in the Petri dish to a synthetic image, in which circular blobs are drawn on a constant background to mark the position of the C. elegans. To apply this simplification method, it is divided into two steps. First, a Faster R-CNN network detects the C. elegans, allowing its transformation into a synthetic image. Second, using the simplified image sequence as input, a regression neural network is in charge of predicting the count of live nematodes on each day of the experiment. In this way, the counting network was trained using a simple simulator, avoiding labeling a very large real dataset or developing a realistic simulator. Results showed that the differences between the curves obtained by the proposed method and the manual curves are not statistically significant for either short-lived N2 (p-value log rank test 0.45) or long-lived daf-2 (p-value log rank test 0.83) strains.

Towards generalization for Caenorhabditis elegans detection.

The nematode Caenorhabditis elegans (C. elegans) is of significant interest for research into neurodegenerative diseases, aging, and drug screening. However, conducting these assays manually is a tedious and time-consuming process. This paper proposes a methodology to achieve a generalist C. elegans detection algorithm, as previous work only focused on dataset-specific detection, tailored exclusively to the characteristics and appearance of the images in a given dataset. The main aim of our study is to achieve a solution that allows for robust detection, regardless of the image-capture system used, with the potential to serve as a basis for the automation of numerous assays. These potential applications include worm counting, worm tracking, motion detection and motion characterization. To train this model, a dataset consisting of a wide variety of appearances adopted by C. elegans has been curated and dataset augmentation methods have been proposed and evaluated, including synthetic image generation. The results show that the model achieves an average precision of 89.5% for a wide variety of C. elegans appearances that were not used during training, thereby validating its generalization capabilities.

Automatic segmentation of Caenorhabditis elegans skeletons in worm aggregations using improved U-Net in low-resolution image sequences.

Pose estimation of C. elegans in image sequences is challenging and even more difficult in low-resolution images. Problems range from occlusions, loss of worm identity, and overlaps to aggregations that are too complex or difficult to resolve, even for the human eye. Neural networks, on the other hand, have shown good results in both low-resolution and high-resolution images. However, training in a neural network model requires a very large and balanced dataset, which is sometimes impossible or too expensive to obtain. In this article, a novel method for predicting C. elegans poses in cases of multi-worm aggregation and aggregation with noise is proposed. To solve this problem we use an improved U-Net model capable of obtaining images of the next aggregated worm posture. This neural network model was trained/validated using a custom-generated dataset with a synthetic image simulator. Subsequently, tested with a dataset of real images. The results obtained were greater than 75% in precision and 0.65 with Intersection over Union (IoU) values.

Analysis of a C. elegans lifespan prediction method based on a bimodal neural network and uncertainty estimation.

In recent decades, assays with the nematode Caenorhabditis elegans (C. elegans) have enabled great advances to be made in research on aging. However, performing these assays manually is a laborious task. To solve this problem, numerous C. elegans assay automation techniques are being developed to increase throughput and accuracy. In this paper, a method for predicting the lifespan of C. elegans nematodes using a bimodal neural network is proposed and analyzed. Specifically, the model uses the sequence of images and the count of live C. elegans up to the current day to predict the lifespan curve termination. This network has been trained using a simulator to avoid the labeling costs of training such a model. In addition, a method for estimating the uncertainty of the model predictions has been proposed. Using this uncertainty, a criterion has been analyzed to decide at what point the assay could be halted and the user could rely on the model's predictions. The method has been analyzed and validated using real experiments. The results show that uncertainty is reduced from the mean lifespan and that most of the predictions obtained do not present statistically significant differences with respect to the curves obtained manually.

Multiview motion tracking based on a cartesian robot to monitor Caenorhabditis elegans in standard Petri dishes.

Data from manual healthspan assays of the nematode Caenorhabditis elegans (C. elegans) can be complex to quantify. The first attempts to quantify motor performance were done manually, using the so-called thrashing or body bends assay. Some laboratories have automated these approaches using methods that help substantially to quantify these characteristic movements in small well plates. Even so, it is sometimes difficult to find differences in motor behaviour between strains, and/or between treated vs untreated worms. For this reason, we present here a new automated method that increases the resolution flexibility, in order to capture more movement details in large standard Petri dishes, in such way that those movements are less restricted. This method is based on a Cartesian robot, which enables high-resolution images capture in standard Petri dishes. Several cameras mounted strategically on the robot and working with different fields of view, capture the required C. elegans visual information. We have performed a locomotion-based healthspan experiment with several mutant strains, and we have been able to detect statistically significant differences between two strains that show very similar movement patterns.

Caenorhabditis elegans Multi-Tracker Based on a Modified Skeleton Algorithm.

Automatic tracking of Caenorhabditis elegans (C. egans) in standard Petri dishes is challenging due to high-resolution image requirements when fully monitoring a Petri dish, but mainly due to potential losses of individual worm identity caused by aggregation of worms, overlaps and body contact. To date, trackers only automate tests for individual worm behaviors, canceling data when body contact occurs. However, essays automating contact behaviors still require solutions to this problem. In this work, we propose a solution to this difficulty using computer vision techniques. On the one hand, a skeletonization method is applied to extract skeletons in overlap and contact situations. On the other hand, new optimization methods are proposed to solve the identity problem during these situations. Experiments were performed with 70 tracks and 3779 poses (skeletons) of C. elegans. Several cost functions with different criteria have been evaluated, and the best results gave an accuracy of 99.42% in overlapping with other worms and noise on the plate using the modified skeleton algorithm and 98.73% precision using the classical skeleton algorithm.

Towards Lifespan Automation for Caenorhabditis elegans Based on Deep Learning: Analysing Convolutional and Recurrent Neural Networks for Dead or Live Classification.

The automation of lifespan assays with C. elegans in standard Petri dishes is a challenging problem because there are several problems hindering detection such as occlusions at the plate edges, dirt accumulation, and worm aggregations. Moreover, determining whether a worm is alive or dead can be complex as they barely move during the last few days of their lives. This paper proposes a method combining traditional computer vision techniques with a live/dead C. elegans classifier based on convolutional and recurrent neural networks from low-resolution image sequences. In addition to proposing a new method to automate lifespan, the use of data augmentation techniques is proposed to train the network in the absence of large numbers of samples. The proposed method achieved small error rates (3.54% ± 1.30% per plate) with respect to the manual curve, demonstrating its feasibility.

Small flexible automated system for monitoring Caenorhabditis elegans lifespan based on active vision and image processing techniques.

Traditionally Caenorhabditis elegans lifespan assays are performed by manually inspecting nematodes with a dissection microscope, which involves daily counting of live/dead worms cultured in Petri plates for 21-25 days. This manual inspection requires the screening of hundreds of worms to ensure statistical robustness, and is therefore a time-consuming approach. In recent years, various automated artificial vision systems have been reported to increase the throughput, however they usually provide less accurate results than manual assays. The main problems identified when using these vision systems are the false positives and false negatives, which occur due to culture media changes, occluded zones, dirtiness or condensation of the Petri plates. In this work, we developed and described a new C. elegans monitoring machine, SiViS, which consists of a flexible and compact platform design to analyse C. elegans cultures using the standard Petri plates seeded with E. coli. Our system uses an active vision illumination technique and different image-processing pipelines for motion detection, both previously reported, providing a fully automated image processing pipeline. In addition, this study validated both these methods and the feasibility of the SiViS machine for lifespan experiments by comparing them with manual lifespan assays. Results demonstrated that the automated system yields consistent replicates (p-value log rank test 0.699), and there are no significant differences between automated system assays and traditionally manual assays (p-value 0.637). Finally, although we have focused on the use of SiViS in longevity assays, the system configuration is flexible and can, thus, be adapted to other C. elegans studies such as toxicity, mobility and behaviour.

Improving skeleton algorithm for helping Caenorhabditis elegans trackers.

One of the main problems when monitoring Caenorhabditis elegans nematodes (C. elegans) is tracking their poses by automatic computer vision systems. This is a challenge given the marked flexibility that their bodies present and the different poses that can be performed during their behaviour individually, which become even more complicated when worms aggregate with others while moving. This work proposes a simple solution by combining some computer vision techniques to help to determine certain worm poses and to identify each one during aggregation or in coiled shapes. This new method is based on the distance transformation function to obtain better worm skeletons. Experiments were performed with 205 plates, each with 10, 15, 30, 60 or 100 worms, which totals 100,000 worm poses approximately. A comparison of the proposed method was made to a classic skeletonisation method to find that 2196 problematic poses had improved by between 22% and 1% on average in the pose predictions of each worm.

Reducing Results Variance in Lifespan Machines: An Analysis of the Influence of Vibrotaxis on Wild-Type Caenorhabditis elegans for the Death Criterion.

Nowadays, various artificial vision-based machines automate the lifespan assays of C. elegans. These automated machines present wider variability in results than manual assays because in the latter worms can be poked one by one to determine whether they are alive or not. Lifespan machines normally use a "dead or alive criterion" based on nematode position or pose changes, without poking worms. However, worms barely move on their last days of life, even though they are still alive. Therefore, a long monitoring period is necessary to observe motility in order to guarantee worms are actually dead, or a stimulus to prompt worm movement is required to reduce the lifespan variability measure. Here, a new automated vibrotaxis-based method for lifespan machines is proposed as a solution to prompt a motion response in all worms cultured on standard Petri plates in order to better distinguish between live and dead individuals. This simple automated method allows the stimulation of all animals through the whole plate at the same time and intensity, increasing the experiment throughput. The experimental results exhibited improved live-worm detection using this method, and most live nematodes (>93%) reacted to the vibration stimulus. This method increased machine sensitivity by decreasing results variance by approximately one half (from ±1 individual error per plate to ±0.6) and error in lifespan curve was reduced as well (from 2.6% to 1.2%).

Improving lifespan automation for Caenorhabditis elegans by using image processing and a post-processing adaptive data filter.

Automated lifespan determination for C. elegans cultured in standard Petri dishes is challenging. Problems include occlusions of Petri dish edges, aggregation of worms, and accumulation of dirt (dust spots on lids) during assays, etc. This work presents a protocol for a lifespan assay, with two image-processing pipelines applied to different plate zones, and a new data post-processing method to solve the aforementioned problems. Specifically, certain steps in the culture protocol were taken to alleviate aggregation, occlusions, contamination, and condensation problems. This method is based on an active illumination system and facilitates automated image sequence analysis, does not need human threshold adjustments, and simplifies the techniques required to extract lifespan curves. In addition, two image-processing pipelines, applied to different plate zones, were employed for automated lifespan determination. The first image-processing pipeline was applied to a wall zone and used only pixel level information because worm size or shape features were unavailable in this zone. However, the second image-processing pipeline, applied to the plate centre, fused information at worm and pixel levels. Simple death event detection was used to automatically obtain lifespan curves from the image sequences that were captured once daily throughout the assay. Finally, a new post-processing method was applied to the extracted lifespan curves to filter errors. The experimental results showed that the errors in automated counting of live worms followed the Gaussian distribution with a mean of 2.91% and a standard deviation of ±12.73% per Petri plate. Post-processing reduced this error to 0.54 ± 8.18% per plate. The automated survival curve incurred an error of 4.62 ± 2.01%, while the post-process method reduced the lifespan curve error to approximately 2.24 ± 0.55%.

Active backlight for automating visual monitoring: An analysis of a lighting control technique for Caenorhabditis elegans cultured on standard Petri plates.

Lifespan and healthspan machines can undergo C. elegans image segmentation errors due to changes in lighting conditions, which produce non-uniform images. Most C. elegans monitoring machines use backlight techniques based on the transparency of both the container and media. Backlight illumination obtains high-contrast images with dark C. elegans and a bright background. However, changes in illumination or media transparency conditions can produce non-uniform images, which are currently alleviated by image processing techniques. Besides, these machines should avoid C. elegans exposure to light as much as possible because light stresses worms, and can even affect their lifespan, mainly when using (1) long exposure times, (2) high intensities or (3) wavelengths that come close to ultraviolet. However, if short exposure of worms to light is required for visual monitoring, then light can also be used as a movement stimulus. In this paper, an active backlight method is analysed. The proposed method consists of controlling the light intensities and wavelengths of an illumination dots matrix with PID regulators. These regulators adapt illumination to some changing conditions. The experimental results shows that this method simplifies the image segmentation problem because it is able to automatically compensate not only changes in media transparency throughout assay days, but also changes in ambient conditions, such as smooth condensation on the lid and light derivatives of the illumination source during its lifetime. In addition, the strategic application of wavelengths could be adapted for the requirements of each assay. For instance, a specific control strategy has been proposed to minimise stress to worms and trying to stimulate C. elegans movement in lifespan assays.

Dual Quaternions as Constraints in 4D-DPM Models for Pose Estimation.

The goal of this research work is to improve the accuracy of human pose estimation using the Deformation Part Model (DPM) without increasing computational complexity. First, the proposed method seeks to improve pose estimation accuracy by adding the depth channel to DPM, which was formerly defined based only on red-green-blue (RGB) channels, in order to obtain a four-dimensional DPM (4D-DPM). In addition, computational complexity can be controlled by reducing the number of joints by taking it into account in a reduced 4D-DPM. Finally, complete solutions are obtained by solving the omitted joints by using inverse kinematics models. In this context, the main goal of this paper is to analyze the effect on pose estimation timing cost when using dual quaternions to solve the inverse kinematics.

Calibration of a trinocular system formed with wide angle lens cameras.

To obtain 3D information of large areas, wide angle lens cameras are used to reduce the number of cameras as much as possible. However, since images are high distorted, errors in point correspondences increase and 3D information could be erroneous. To increase the number of data from images and to improve the 3D information, trinocular sensors are used. In this paper a calibration method for a trinocular sensor formed with wide angle lens cameras is proposed. First pixels locations in the images are corrected using a set of constraints which define the image formation in a trinocular system. When pixels location are corrected, lens distortion and trifocal tensor is computed.

Accurate calibration with highly distorted images.

Camera calibration is a two-step process where first a linear algebraic approximation is followed by a nonlinear minimization. The nonlinear minimization adjusts the pin-hole and lens distortion models to the calibrating data. Since both models are coupled, nonlinear minimization can converge to a local solution easily. Moreover, nonlinear minimization is poorly conditioned since parameters with different effects in the minimization function are calculated simultaneously (some are in pixels, some in world coordinates, and some are lens distortion parameters). A local solution is adapted to parameters, which minimize the function easily, and the remaining parameters are just adapted to this solution. We propose a calibration method where traditional calibration steps are inverted. First, a nonlinear minimization is done, and after, camera parameters are computed in a linear step. Using projective geometry constraints in a nonlinear minimization process, detected point locations in the images are corrected. The pin-hole and lens distortion models are computed separately with corrected point locations. The proposed method avoids the coupling between both models. Also, the condition of nonlinear minimization increases since points coordinates are computed alone.

Calibration of a wide angle stereoscopic system.

Inaccuracies in the calibration of a stereoscopic system appear with errors in point correspondences between both images and inexact points localization in each image. Errors increase if the stereoscopic system is composed of wide angle lens cameras. We propose a technique where detected points in both images are corrected before estimating the fundamental matrix and the lens distortion models. Since points are corrected first, errors in point correspondences and point localization are avoided. To correct point location in both images, geometrical and epipolar constraints are imposed in a nonlinear minimization problem. Geometrical constraints define the point localization in relation to its neighbors in the same image, and eipolar constraints represent the location of one point referred to its corresponding point in the other image.

Camera calibration under optimal conditions.

Different methods based on photogrammetry or self-calibration exist to calibrate intrinsic and extrinsic camera parameters and also for data pre- and post-processing. From a practical viewpoint, it is quite difficult to decide which calibration method gives accurate results and even whether any data processing is necessary. This paper proposes a set of optimal conditions to resolve the calibration process accurately. The calibration method uses several images of a 2D pattern. Optimal conditions define the number of points and the number of images to resolve the calibration accurately, as well as positions and orientations from where images should be taken.

Lens distortion models evaluation.

Many lens distortion models exist with several variations, and each distortion model is calibrated by using a different technique. If someone wants to correct lens distortion, choosing the right model could represent a very difficult task. Calibration depends on the chosen model, and some methods have unstable results. Normally, the distortion model containing radial, tangential, and prism distortion is used, but it does not represent high distortion accurately. The aim of this paper is to compare different lens distortion models to define the one that obtains better results under some conditions and to explore if some model can represent high and low distortion adequately. Also, we propose a calibration technique to calibrate several models under stable conditions. Since performance is hard conditioned with the calibration technique, the metric lens distortion calibration method is used to calibrate all the evaluated models.