Reliability and accuracy of an automated tracking algorithm to measure controlled passive and active muscle fascicle length changes from ultrasound

Manual tracking of muscle fascicle length changes from ultrasound images is a subjective and time-consuming process. The purpose of this study was to assess the repeatability and accuracy of an automated algorithm for tracking fascicle length changes in the medial gastrocnemius (MG) muscle during passive length changes and active contractions (isometric, concentric and eccentric) performed on a dynamometer. The freely available, automated tracking algorithm was based on the Lucas–Kanade optical flow algorithm with an affine optic flow extension, which accounts for image translation, dilation, rotation and shear between consecutive frames of an image sequence. Automated tracking was performed by three experienced assessors, and within- and between-examiner repeatability was computed using the coefficient of multiple determination (CMD). Fascicle tracking data were also compared with manual digitisation of the same image sequences, and the level of agreement between the two methods was calculated using the coefficient of multiple correlation (CMC). The CMDs across all test conditions ranged from 0.50 to 0.93 and were all above 0.98 when recomputed after the systematic error due to the estimate of the initial fascicle length on the first ultrasound frame was removed from the individual fascicle length waveforms. The automated and manual tracking approaches produced similar fascicle length waveforms, with an overall CMC of 0.88, which improved to 0.94 when the initial length offset was removed. Overall results indicate that the automated fascicle tracking algorithm was a repeatable, accurate and time-efficient method for estimating fascicle length changes of the MG muscle in controlled passive and active conditions.     

Assessing Mitochondrial Movement Within Neurons: Manual Versus Automated Tracking Methods

Owing to the small size of mitochondria and the complexity of their motility patterns, mitochondrial tracking is technically challenging. Mitochondria are often tracked manually; however, this is time‐consuming and prone to measurement error. Here, we examined the suitability of four commercial and open‐source software alternatives for automated mitochondrial tracking in neurons compared with manual measurements. We show that all the automated tracking tools dramatically underestimated track length, mitochondrial displacement and movement duration, with reductions ranging from 45 to 77% of the values obtained manually. In contrast, mitochondrial velocity was generally overestimated. Only the number of motile mitochondria and their directionality were similar between strategies. Despite these discrepancies, we show that automated tools successfully detected transport alterations after applying an oxidant agent. Thus, automated methods appear to be suitable for assessing relative transport differences between experimental groups, but not for absolute quantification of mitochondrial dynamics. Although useful for objective and time‐efficient measurements of mitochondrial movements, results provided by automated methods should be interpreted with caution.      

Computer-assisted tracking of actin filament motility.

In vitro motility assays, in which fluorescently labeled actin filaments are propelled by myosin molecules adhered to a glass coverslip, require that actin filament velocity be determined. We have developed a computer-assisted filament tracking system that reduced the analysis time, minimized investigator bias, and provided greater accuracy in locating actin filaments in video images. The tracking routine successfully tracked filaments under experimental conditions where filament density, size, and extent of photobleaching varied dramatically. Videotaped images of actin filament motility were digitized and processed to enhance filament image contrast relative to background. Once processed, filament images were cross correlated between frames and a filament path was determined. The changes in filament centroid or center position between video frames were then used to calculate filament velocity. The tracking routine performance was evaluated and the sources of noise that contributed to errors in velocity were identified and quantified. Errors originated in algorithms for filament centroid determination and in the choice of sampling interval between video frames. With knowledge of these error sources, the investigator can maximize the accuracy of the velocity calculation through access to user-definable computer program parameters.     

A novel validation algorithm allows for automated cell tracking and the extraction of biologically meaningful parameters

Automated microscopy is currently the only method to non-invasively and label-free observe complex multi-cellular processes, such as cell migration, cell cycle, and cell differentiation. Extracting biological information from a time-series of micrographs requires each cell to be recognized and followed through sequential microscopic snapshots. Although recent attempts to automatize this process resulted in ever improving cell detection rates, manual identification of identical cells is still the most reliable technique. However, its tedious and subjective nature prevented tracking from becoming a standardized tool for the investigation of cell cultures. Here, we present a novel method to accomplish automated cell tracking with a reliability comparable to manual tracking. Previously, automated cell tracking could not rival the reliability of manual tracking because, in contrast to the human way of solving this task, none of the algorithms had an independent quality control mechanism; they missed validation. Thus, instead of trying to improve the cell detection or tracking rates, we proceeded from the idea to automatically inspect the tracking results and accept only those of high trustworthiness, while rejecting all other results. This validation algorithm works independently of the quality of cell detection and tracking through a systematic search for tracking errors. It is based only on very general assumptions about the spatiotemporal contiguity of cell paths. While traditional tracking often aims to yield genealogic information about single cells, the natural outcome of a validated cell tracking algorithm turns out to be a set of complete, but often unconnected cell paths, i.e. records of cells from mitosis to mitosis. This is a consequence of the fact that the validation algorithm takes complete paths as the unit of rejection/acceptance. The resulting set of complete paths can be used to automatically extract important biological parameters with high reliability and statistical significance. These include the distribution of life/cycle times and cell areas, as well as of the symmetry of cell divisions and motion analyses. The new algorithm thus allows for the quantification and parameterization of cell culture with unprecedented accuracy. To evaluate our validation algorithm, two large reference data sets were manually created. These data sets comprise more than 320,000 unstained adult pancreatic stem cells from rat, including 2592 mitotic events. The reference data sets specify every cell position and shape, and assign each cell to the correct branch of its genealogic tree. We provide these reference data sets for free use by others as a benchmark for the future improvement of automated tracking methods.     

Automated counting of phytoplankton by pattern recognition: a comparison with a manual counting method

Computer-based image analysis and pattern recognition methodswere used to construct a system able automatically to identify,count and measure selected groups of phytoplankton. An imageanalysis algorithm was employed to isolate and measure objectsfrom digitized images of a phytoplankton sample. The measurementsobtained were used to identify selected groups of phytoplanktonby a combination of artificial neural networks and simple rule-basedprocedures. The system was trained and tested using samplesof lake water covering an annual growth cycle from Lough Neaghin Northern Ireland. Total volume estimates were obtained forthe four major phytoplankton species, using both the automatedsystem and a manual counting method. Estimates of total cellvolume obtained from the automated system were within 10% ofthose derived by manual analysis of the same cells. The automatedsystem produced total cell volume estimates close to those obtainedfrom manual analysis of different aliquots of the same watersample. Variation between successive counts of the same watersample was higher with the automated system than with the manualcounting method. Limitations and possible improvements to thetechnology are discussed.     

Tracking single proteins within cells

We present experiments in which single proteins were imaged and tracked within mammalian cells. Single proteins of R-phycoerythrin (RPE) were imaged by epifluorescence microscopy in the nucleoplasm and cytoplasm at 71 frames/s. We acquired two-dimensional trajectories of proteins (corresponding to the projection of three-dimensional trajectories onto the plane of focus) for an average of 17 frames in the cytoplasm and 16 frames in the nucleus. Diffusion constants were determined from linear fits to the mean square displacement and from the mean displacement squared per frame. We find that the distribution of diffusion constants for RPE within cells is broader than the distributions obtained from RPE in a glycerol solution, from a Monte Carlo simulation, and from the theoretical distribution for simple diffusion. This suggests that on the time scales of our measurements, the motion of single RPE proteins in the cytoplasm and nucleoplasm cannot be modeled by simple diffusion with a unique diffusion constant. Our results demonstrate that it is possible to follow the motion of single proteins within cells and that the technique of single molecule tracking can be used to probe the dynamics of intracellular macromolecules.     

Automated Cell Tracking and Analysis in Phase-Contrast Videos (iTrack4U): Development of Java Software Based on Combined Mean-Shift Processes

Cell migration is a key biological process with a role in both physiological and pathological conditions. Locomotion of cells during embryonic development is essential for their correct positioning in the organism; immune cells have to migrate and circulate in response to injury. Failure of cells to migrate or an inappropriate acquisition of migratory capacities can result in severe defects such as altered pigmentation, skull and limb abnormalities during development, and defective wound repair, immunosuppression or tumor dissemination. The ability to accurately analyze and quantify cell migration is important for our understanding of development, homeostasis and disease. In vitro cell tracking experiments, using primary or established cell cultures, are often used to study migration as cells can quickly and easily be genetically or chemically manipulated. Images of the cells are acquired at regular time intervals over several hours using microscopes equipped with CCD camera. The locations (x,y,t) of each cell on the recorded sequence of frames then need to be tracked. Manual computer-assisted tracking is the traditional method for analyzing the migratory behavior of cells. However, this processing is extremely tedious and time-consuming. Most existing tracking algorithms require experience in programming languages that are unfamiliar to most biologists. We therefore developed an automated cell tracking program, written in Java, which uses a mean-shift algorithm and ImageJ as a library. iTrack4U is a user-friendly software. Compared to manual tracking, it saves considerable amount of time to generate and analyze the variables characterizing cell migration, since they are automatically computed with iTrack4U. Another major interest of iTrack4U is the standardization and the lack of inter-experimenter differences. Finally, iTrack4U is adapted for phase contrast and fluorescent cells.     

Continuous thickness measurement of rectus femoris muscle in ultrasound image sequences: A completely automated approach

Muscle thickness is one of the most widely used parameters for quantifying muscle function in both diagnosis and rehabilitation assessment. Ultrasound imaging has been frequently used to non-invasively study the thickness of human muscles as a reliable method. However, the measurement is traditionally conducted by manual digitization of reference points at the superior and inferior muscle fascias, thus it is subjective and time-consuming. In this paper, a novel method is proposed to detect the muscle thickness automatically. The superficial and deep fascias of a muscle are detected by line detection algorithm at the first ultrasound frame, and the image regions of interest (ROI) for the fascias are subsequently located and tracked by optical flow technique. The muscle thickness is geometrically obtained based on the location of the fascias for each frame. Six ultrasound sequences (250 frames in each sequence) are used to evaluate this method. The correlation coefficient of the detection results between the proposed method and manual method is 0.95 ± 0.01, and the difference is ?0.05 ± 0.22 mm. The linear regression of the total 1500 detections show that a good linear correlation between the results of the two methods is obtained (R² = 0.981). The automated method proposed here provides an accurate, high repeatable and efficient approach for estimating fascicle thickness during human motion, thus justifying its application in biological sciences.     

An algorithm for automated analysis of ultrasound images to measure tendon excursion in vivo

The accuracy of an algorithm for the automated tracking of tendon excursion from ultrasound images was tested in three experiments. Because the automated method could not be tested against direct measurements of tendon excursion in vivo, an indirect validation procedure was employed. In one experiment, a wire "phantom" was moved a known distance across the ultrasound probe and the automated tracking results were compared with the known distance. The excursion of the musculotendinous junction of the gastrocnemius during frontal and sagittal plane movement of the ankle was assessed in a single cadaver specimen both by manual tracking and with a cable extensometer sutured to the gastrocnemius muscle. A third experiment involved estimation of Achilles tendon excursion in vivo with both manual and automated tracking. Root mean squared (RMS) error was calculated between pairs of measurements after each test. Mean RMS errors of less than 1 mm were observed for the phantom experiments. For the in vitro experiment, mean RMS errors of 8-9% of the total tendon excursion were observed. Mean RMS errors of 6-8% of the total tendon excursion were found in vivo. The results indicate that the proposed algorithm accurately tracks Achilles tendon excursion, but further testing is necessary to determine its general applicability.     

Optimal Geometrical Set for Automated Marker Placement to Virtualized Real-Time Facial Emotions

In recent years, real-time face recognition has been a major topic of interest in developing intelligent human-machine interaction systems. Over the past several decades, researchers have proposed different algorithms for facial expression recognition, but there has been little focus on detection in real-time scenarios. The present work proposes a new algorithmic method of automated marker placement used to classify six facial expressions: happiness, sadness, anger, fear, disgust, and surprise. Emotional facial expressions were captured using a webcam, while the proposed algorithm placed a set of eight virtual markers on each subject’s face. Facial feature extraction methods, including marker distance (distance between each marker to the center of the face) and change in marker distance (change in distance between the original and new marker positions), were used to extract three statistical features (mean, variance, and root mean square) from the real-time video sequence. The initial position of each marker was subjected to the optical flow algorithm for marker tracking with each emotional facial expression. Finally, the extracted statistical features were mapped into corresponding emotional facial expressions using two simple non-linear classifiers, K-nearest neighbor and probabilistic neural network. The results indicate that the proposed automated marker placement algorithm effectively placed eight virtual markers on each subject’s face and gave a maximum mean emotion classification rate of 96.94% using the probabilistic neural network.     

Automated detection and tracking of individual and clustered cell surface low density lipoprotein receptor molecules.

We have developed a technique to detect, recognize, and track each individual low density lipoprotein receptor (LDL-R) molecule and small receptor clusters on the surface of human skin fibroblasts. Molecular recognition and high precision (30 nm) simultaneous automatic tracking of all of the individual receptors in the cell surface population utilize quantitative time-lapse low light level digital video fluorescence microscopy analyzed by purpose-designed algorithms executed on an image processing work station. The LDL-Rs are labeled with the biologically active, fluorescent LDL derivative dil-LDL. Individual LDL-Rs and unresolved small clusters are identified by measuring the fluorescence power radiated by the sub-resolution fluorescent spots in the image; identification of single particles is ascertained by four independent techniques. An automated tracking routine was developed to track simultaneously, and without user intervention, a multitude of fluorescent particles through a sequence of hundreds of time-lapse image frames. The limitations on tracking precision were found to depend on the signal-to-noise ratio of the tracked particle image and mechanical drift of the microscope system. We describe the methods involved in (i) time-lapse acquisition of the low-light level images, (ii) simultaneous automated tracking of the fluorescent diffraction limited punctate images, (iii) localizing particles with high precision and limitations, and (iv) detecting and identifying single and clustered LDL-Rs. These methods are generally applicable and provide a powerful tool to visualize and measure dynamics and interactions of individual integral membrane proteins on living cell surfaces.     

ENUMERATING 3MTM PETRIFILMTM AEROBIC COUNT PLATES USING THE PETRISCAN® AUTOMATED COLONY COUNTER

In the food and dairy industries, aerobic plate counts are determined by a time-consuming and laborious hand-counting method. The PetriScan ® automated colony counter was developed to improve efficiency in the microbiology laboratory. In this study, colony counts of food, dairy, and milk products plated on 3M^TM Petrifilm^TM Aerobic Count Plates were compared using both automated and manual count plate methods. For sample variation, 16 different food, dairy, and milk products were used. Samples were prepared and serially diluted using Butterfield's diluent according to approved AOAC methods and APHA's Standard Methods. Plates were inoculated, incubated, and counted according to AOAC methods. For data collection, plates with counts between 5 and 300 colonies were included. A total of 55 low (5–30), 29 medium (31–100), and 23 high (101–300) count plates were used. Duplicate results were recorded for both methods; hand counts were tallied by two scientists. The duplicates of the mean log values for manual counts varied by 0.0005 and 0.0007, and the duplicates for the automated counts varied by 0.0011. The mean log value difference between the automated and manual counts for pooled data was 0.035. The correlation coefficient for the regression line comparing the automated and manual count methods for pooled data was 0.98. The regression equation was y = 0.9257x + 0.0781. These results demonstrate that the PetriScan® automated colony counter is a comparable and practical alternative to the standard method of manually counting plates.     

Comparison of glenohumeral joint kinematics between manual wheelchair tasks and implications on the subacromial space: A biplane fluoroscopy study

Scapula and humerus motion associated with common manual wheelchair tasks is hypothesized to reduce the subacromial space. However, previous work relied on either marker-based motion capture for kinematic measures, which is prone to skin-motion artifact; or ultrasound imaging for arthrokinematic measures, which are 2D and acquired in statically-held positions. The aim of this study was to use a fluoroscopy-based approach to accurately quantify glenohumeral kinematics during manual wheelchair use, and compare tasks for a subset of parameters theorized to be associated with mechanical impingement. Biplane images of the dominant shoulder were acquired during scapular plane elevation, propulsion, sideways lean, and weight-relief raise in ten manual wheelchair users with spinal cord injury. A computed tomography scan of the shoulder was obtained, and model-based tracking was used to quantify six-degree-of-freedom glenohumeral kinematics. Axial rotation and superior/inferior and anterior/posterior humeral head positions were characterized for full activity cycles and compared between tasks. The change in the subacromial space was also determined for the period of each task defined by maximal change in the aforementioned parameters. Propulsion, sideways lean, and weight-relief raise, but not scapular plane elevation, were marked by mean internal rotation (8.1°, 10.8°, 14.7°, −49.2° respectively). On average, the humeral head was most superiorly positioned during the weight-relief raise (1.6 ± 0.9 mm), but not significantly different from the sideways lean (0.8 ± 1.1 mm) (p = 0.191), and much of the task was characterized by inferior translation. Scaption was the only task without a defined period of superior translation on average. Pairwise comparisons revealed no significant differences between tasks for anterior/posterior position (task means range: 0.1–1.7 mm), but each task exhibited defined periods of anterior translation. There was not a consistent trend across tasks between internal rotation, superior translation, and anterior translation with reductions in the subacromial space. Further research is warranted to determine the likelihood of mechanical impingement during these tasks based on the measured task kinematics and reductions in the subacromial space.     

Automatic tracking of medial gastrocnemius fascicle length during human locomotion

During human locomotion lower extremity muscle-tendon units undergo cyclic length changes that were previously assumed to be representative of muscle fascicle length changes. Measurements in cats and humans have since revealed that muscle fascicle length changes can be uncoupled from those of the muscle-tendon unit. Ultrasonography is frequently used to estimate fascicle length changes during human locomotion. Fascicle length analysis requires time consuming manual methods that are prone to human error and experimenter bias. To bypass these limitations, we have developed an automatic fascicle tracking method based on the Lucas-Kanade optical flow algorithm with an affine optic flow extension. The aims of this study were to compare gastrocnemius fascicle length changes during locomotion using the automated and manual approaches and to determine the repeatability of the automated approach. Ultrasound was used to examine gastrocnemius fascicle lengths in eight participants walking at 4, 5, 6, and 7 km/h and jogging at 7 km/h on a treadmill. Ground reaction forces and three dimensional kinematics were recorded simultaneously. The level of agreement between methods and the repeatability of the automated method were quantified using the coefficient of multiple correlation (CMC). Regardless of speed, the level of agreement between methods was high, with overall CMC values of 0.90 ± 0.09 (95% CI: 0.86-0.95). Repeatability of the algorithm was also high, with an overall CMC of 0.88 ± 0.08 (95% CI: 0.79-0.96). The automated fascicle tracking method presented here is a robust, reliable, and time-efficient alternative to the manual analysis of muscle fascicle length during gait.     

An agent-based system for re-annotation of genomes

Genome annotation projects can produce incorrect results if they are based on obsolete data or inappropriate models. We have developed an automatic re-annotation system that uses agents to perform repetitive tasks and reports the results to the user. These tasks involve BLAST searches on biological databases (GenBank) and the use of detection tools (Genemark and Glimmer) to identify new open reading frames. Several agents execute these tools and combine their results to produce a list of open reading frames that is sent back to the user. Our goal was to reduce the manual work, executing most tasks automatically by computational tools. A prototype was implemented and validated using Mycoplasma pneumoniae and Haemophilus influenzae original annotated genomes. The results reported by the system identify most of new features present in the re-annotated versions of these genomes.     

Strategies for rare-event detection: an approach for automated fetal cell detection in maternal blood.

This article explores the feasibility of the use of automated microscopy and image analysis to detect the presence of rare fetal nucleated red blood cells (NRBCs) circulating in maternal blood. The rationales for enrichment and for automated image analysis for "rare-event" detection are reviewed. We also describe the application of automated image analysis to 42 maternal blood samples, using a protocol consisting of one-step enrichment followed by immunocytochemical staining for fetal hemoglobin (HbF) and FISH for X- and Y-chromosomal sequences. Automated image analysis consisted of multimode microscopy and subsequent visual evaluation of image memories containing the selected objects. The FISH results were compared with the results of conventional karyotyping of the chorionic villi. By use of manual screening, 43% of the slides were found to be positive (>=1 NRBC), with a mean number of 11 NRBCs (range 1-40). By automated microscopy, 52% were positive, with on average 17 NRBCs (range 1-111). There was a good correlation between both manual and automated screening, but the NRBC yield from automated image analysis was found to be superior to that from manual screening (P=.0443), particularly when the NRBC count was >15. Seven (64%) of 11 XY fetuses were correctly diagnosed by FISH analysis of automatically detected cells, and all discrepancies were restricted to the lower cell-count range. We believe that automated microscopy and image analysis reduce the screening workload, are more sensitive than manual evaluation, and can be used to detect rare HbF-containing NRBCs in maternal blood.     

Motion representation of the long fingers: A proposal for the definitions of new anatomical frames

Despite the availability of the International Society of Biomechanics (ISB) recommendations for the orientation of anatomical frames, no consensus exists about motion representations related to finger kinematics. This paper proposes novel anatomical frames for motion representation of the phalangeal segments of the long fingers. A three-dimensional model of a human forefinger was acquired from a non-pathological fresh-frozen hand. Medical imaging was used to collect phalangeal discrete positions. Data processing was performed using a customized software interface (“lhpFusionBox”) to create a specimen-specific model and to reconstruct the discrete motion path. Five examiners virtually palpated two sets of landmarks. These markers were then used to build anatomical frames following two methods: a reference method following ISB recommendations and a newly-developed method based on the mean helical axis (HA). Motion representations were obtained and compared between examiners. Virtual palpation precision was around 1 mm, which is comparable to results from the literature. The comparison of the two methods showed that the helical axis method seemed more reproducible between examiners especially for secondary, or accessory, motions. Computed Root Mean Square distances comparing methods showed that the ISB method displayed a variability 10 times higher than the HA method. The HA method seems to be suitable for finger motion representation using discrete positions from medical imaging. Further investigations are required before being able to use the methodology with continuous tracking of markers set on the subject?s hand.     

Optimal marker set assessment for motion capture of 3D mimic facial movements

Nowadays, facial mimicry studies have acquired a great importance in the clinical domain and 3D motion capture systems are becoming valid tools for analysing facial muscles movements, thanks to the remarkable developments achieved in the 1990s.However, the face analysis domain suffers from a lack of valid motion capture protocol, due to the complexity of the human face. Indeed, a framework for defining the optimal marker set layout does not exist yet and, up to date, researchers still use their traditional facial point sets with manually allocated markers.Therefore, the study proposes an automatic approach to compute a minimum optimized marker layout to be exploited in facial motion capture, able to simplify the marker allocation without decreasing the significance level. Specifically, the algorithm identifies the optimal facial marker layouts selecting the subsets of linear distances among markers that allow to automatically recognizing with the highest performances, through a k-nearest neighbours classification technique, the acted facial movements. The marker layouts are extracted from them. Various validation and testing phases have demonstrated the accuracy, robustness and usefulness of the custom approach.