SleepNet: automated sleep analysis via dense convolutional neural network using physiological time series.

OBJECTIVE: In this work, a dense recurrent convolutional neural network (DRCNN) was constructed to detect sleep disorders including arousal, apnea and hypopnea using polysomnography (PSG) measurement channels provided in the 2018 PhysioNet Challenge database. APPROACH: Our model structure is composed of multiple dense convolutional units (DCU) followed by a bidirectional long-short term memory (LSTM) layer followed by a softmax output layer. The sleep events, including sleep stages, arousal regions and multiple types of apnea and hypopnea, are manually annotated by experts, which enables us to train our proposed network using a multi-task learning mechanism. Three binary cross-entropy loss functions, corresponding to sleep/wake, target arousal and apnea-hypopnea/normal detection tasks, are summed up to generate our overall network loss function that is optimized using the Adam method. Our model performance was evaluated using two metrics: the area under the precision-recall curve (AUPRC) and the area under the receiver operating characteristic curve (AUROC). To measure our model generalization, 4-fold cross-validation was also performed. For training, our model was applied to full night recording data. MAIN RESULTS: Finally, the average AUPRC and AUROC values associated with the arousal detection task were 0.505 and 0.922, respectively, on our testing dataset. An ensemble of four models trained on different data folds improved the AUPRC and AUROC to 0.543 and 0.931, respectively. SIGNIFICANCE: Our proposed algorithm achieved the first place in the official stage of the 2018 PhysioNet Challenge for detecting sleep arousals with an AUPRC of 0.54 on the blind testing dataset.

Aerobic respiratory costs of swimming in the negatively buoyant brief squid Lolliguncula brevis.

Because of the inherent inefficiency of jet propulsion, squid are considered to be at a competitive disadvantage compared with fishes, which generally depend on forms of undulatory/oscillatory locomotion. Some squid, such as the brief squid Lolliguncula brevis, swim at low speeds in shallow-water complex environments, relying heavily on fin activity. Consequently, their swimming costs may be lower than those of the faster, more pelagic squid studied previously and competitive with those of ecologically relevant fishes. To examine aerobic respiratory swimming costs, O(2) consumption rates were measured for L. brevis of various sizes (2-9 cm dorsal mantle length, DML) swimming over a range of speeds (3-30 cm s(-1)) in swim tunnel respirometers, while their behavior was videotaped. Using kinematic data from swimming squid and force data from models, power curves were also generated. Many squid demonstrated partial (J-shaped) or full (U-shaped) parabolic patterns of O(2) consumption rate as a function of swimming speed, with O(2) consumption minima at 0.5-1.5 DML s(-1). Power curves derived from hydrodynamic data plotted as a function of swimming speed were also parabolic, with power minima at 1.2-1.7 DML s(-1). The parabolic relationship between O(2) consumption rate/power and speed, which is also found in aerial flyers such as birds, bats and insects but rarely in aquatic swimmers because of the difficulties associated with low-speed respirometry, is the result of the high cost of generating lift and maintaining stability at low speeds and overcoming drag at high speeds. L. brevis has a lower rate of O(2) consumption than the squid Illex illecebrosus and Loligo opalescens studied in swim tunnel respirometers and is energetically competitive (especially at O(2) consumption minima) with fishes, such as striped bass, mullet and flounder. Therefore, the results of this study indicate that, like aerial flyers, some negatively buoyant nekton have parabolic patterns of O(2) consumption rate/power as a function of speed and that certain shallow-water squid using considerable fin activity have swimming costs that are competitive with those of ecologically relevant fishes.

Swimming mechanics and behavior of the shallow-water brief squid Lolliguncula brevis.

Although squid are among the most versatile swimmers and rely on a unique locomotor system, little is known about the swimming mechanics and behavior of most squid, especially those that swim at low speeds in inshore waters. Shallow-water brief squid Lolliguncula brevis, ranging in size from 1.8 to 8.9 cm in dorsal mantle length (DML), were placed in flumes and videotaped, and the data were analyzed using motion-analysis equipment. Flow visualization and force measurement experiments were also performed in water tunnels. Mean critical swimming speeds (U(crit)) ranged from 15.3 to 22.8 cm s(-1), and mean transition speeds (U(t); the speed above which squid swim exclusively in a tail-first orientation) varied from 9.0 to 15.3 cm s(-1). At low speeds, negatively buoyant brief squid generated lift and/or improved stability by positioning the mantle and arms at high angles of attack, directing high-speed jets downwards (angles >50 degrees ) and using fin activity. To reduce drag at high speeds, the squid decreased angles of attack and swam tail-first. Fin motion, which could not be characterized exclusively as drag- or lift-based propulsion, was used over 50-95 % of the sustained speed range and provided as much as 83.8 % of the vertical and 55.1 % of the horizontal thrust. Small squid (<3.0 cm DML) used different swimming strategies from those of larger squid, possibly to maximize thrust benefits from vortex ring formation. Furthermore, brief squid employed various unsteady behaviors, such as manipulating funnel diameter during jetting, altering arm position and swimming in different orientations, to boost swimming performance. These results demonstrate that locomotion in slow-swimming squid is complex, involving intricate spatial and temporal interactions between the mantle, fins, arms and funnel.

A mass transfer explanation of metabolic scaling relations in some aquatic invertebrates and algae.

Chemical engineering theory can be used in accounting for the broad range of metabolic scaling exponents found in some aquatic invertebrates and algae. Delivery of metabolically important compounds to these organisms occurs by diffusion through a boundary layer. Dimensionless relations (Sherwood-Reynolds number functions) demonstrate the degree to which water motion and organism size affect mass transfer, and ultimately, metabolic rate. Derivation of mass exponents in the range 0.31 to 1.25 for simple geometries such as plates, spheres, and cylinders directly follows from knowledge of the Sherwood-Reynolds number relations. The range of exponents predicted is that found by allometric studies of metabolic rate in these organisms.

Fluid dynamics in suspension-feeding blackfish.

Measurements of flow patterns and water velocities inside the oral cavity of blackfish (Orthodon microlepidotus), made with a fiberoptic endoscope and thermistor flow probe, revealed that gill-arch structures act in blackfish as barriers that direct particle-laden water to the mucus-covered roof of the oral cavity, where particles are retained. Gill-arch structures have previously been assumed to be the site of particle retention in suspension-feeding fishes. Water does not pass between these structures in blackfish, and they do not serve as filters that separate particles from the water. These results emphasize the importance of directly assessing flow velocity and direction inside the oral cavity of vertebrate suspension feeders, particularly at the level of the filtering elements.

Passive Suspension Feeding by an Octocoral in Plankton Patches: Empirical Test of a Mathematical Model.

Feeding rate in the octocoral, Alcyonium siderium, was investigated as a function of colony size, flow speed, and prey concentration. The feeding rate decreases with time in high prey concentrations. A model of passive suspension feeding is formulated that successfully predicts feeding behavior. At low prey concentrations, the model predicts a linear feeding response as particle flux or colony size increases. The dominant constraint on feeding is the "handling time" required to transfer prey from tentacle to pharynx and to re-extend the tentacle. The time constant of prey capture shows no relation to particle flux, in agreement with the model. Another constraint, the "filtration time," is inversely related to colony size and flow speed. Filtration time becomes important only during feeding in sparse prey concentrations, when feeding rate is proportional to flow speed, colony size, and prey concentration. In the field, Alcyonium colonies reduce filtration time by orienting at right angles to the dominant flow direction. Feeding efficiency on prey patches is low and inversely related to flow speed, colony size, and prey concentration. Feeding in patches is not a simple process for this octocoral, because colonies will "saturate" with prey before all polyps have successfully captured a single prey item.

The Effects of Flow on Polyp-Level Prey Capture in an Octocoral, Alcyonium siderium.

Particle capture by individual polyps and tentacles of the octocoral, Alcyonium siderium, was investigated in flows of different speed and turbulence intensity. In low flow (Umean = 2.7 cm/s; u' = 1.2 cm/s, where u' is the root mean square of the fluctuations from Umean), tentacles on the upstream side of a polyp capture the most prey. In intermediate flow (Umean = 12.2 cm/s; u' = 6.0 cm/s), downstream tentacles within a polyp catch the most prey. In high flow (Umean = 19.8 cm/s; u'= 4.0 cm/s), polyps are bent downstream, eddies form over the tentacular surfaces, and the capture distribution over tentacles becomes radially symmetric. At all flow speeds tested, particles are caught with increasing frequency nearer the tip of the tentacle relative to locations near the pharynx. At the highest flow speed tested, no particles are caught on the segment of each tentacle closest to the pharynx. The per polyp capture efficiency is low and drops markedly with increasing Reynolds number. The capture mechanism for this species appears to be direct interception; inertial impaction is shown to be unimportant. Flow modulation of particle capture by polyps is probably a general phenomenon among octocorals.

Forced convection modulates gas exchange in cnidarians.

Boundary layer thickness is a potentially important component of the diffusive pathway for gas exchange in aquatic organisms. The soft coral Alcyonium siderium (Octocorallia) and sea anemone Metridium senile (Actiniaria) exhibit significant increases in respiration with water flow over a range of Reynolds numbers encountered subtidally. A nondimensional mass transfer analysis of the effect of forced convection demonstrates the importance of the state of the organism's boundary layer in regulating metabolism in these invertebrates. Flow-modulated gas exchange may limit secondary productivity in subtidal environments.

Plasminogen activator: the major secreted neutral protease of cultured skeletal muscle cells.

Clonal mouse skeletal muscle cells which differentiate in culture and form synapses with neuronal cells were found to secrete high levels of protease activity as measured with an 125I-fibrin assay. The secreted proteolytic activity was more than 90% dependent upon the presence of plasminogen in the medium, and had a pH optimum at 7 to 8. This activity was not inhibited by n-ethylmaleimide, pepstatin, EDTA, or EGTA. At millimolar concentrations, greater than 90% inhibition was obtained with either soybean trypsin inhibitor, epsilon aminocaproic acid, Trasylol, or leupeptin. Almost complete inhibition occurred with 1 mM diisopropylfluorophosphate suggesting the presence of a serine residue at the catalytic site. In contrast to the high levels of secreted activity, a lower steady-state level of cell-associated protease activity was detected in cell lysates. The high level of plasminogen activator secreted into the medium of cultured muscle cells suggests a role for such extracellular protease activity in myogenesis during development and remodeling following muscle injury. Such information may be useful in understanding the initial degeneration of neuromuscular contacts in experimental and pathologic denervation.

Noncontact profiling techniques for intraocular lens edge measurements.

Two techniques, focal spot projection and fringe projection, have been developed for profiling the edges of anterior-chamber style intraocular lenses (IOL). The techniques are based on measurement of the trajectory of reflected light when an incident beam is focused and stepped across the edge surface (spot projection) or when two focused beams combine to form an optical interference pattern on the edge (fringe projection). The optical configurations for performing the projection techniques are described along with the mathematical relationship between the reflected-beam trajectories and the edge profile. An automated system for the implementation of focal spot projection was developed. The system has the capability of positioning the IOL edge at the focal point of the incident beam, acquiring trajectory data via stepper motors and a TV camera, processing the data to generate a profile, and displaying the profile on the computer video monitor or in hard-copy form via a printer. On one lens tested, three profiles corresponding to different locations along the IOL edge could be generated without user intervention. However, the user has the option of manually positioning the IOL to permit greater versatility. The typical 300-microm wide IOL edges are profiled to a resolution of approximately 20 microm. To constrain optical and microprocessor hardware costs, compromises were made in operating speeds so that approximately 20 min is required per profile. With current trends in integrated circuit technology future generations of the system could be made faster and prove useful as a QC tool for the IOL industry. Furthermore, the present device can be used in determining appropriate IOL profile standards.

Neurotrophic control of 16S acetylcholinesterase from mammalian skeletal muscle in organ culture.

The effects of rat obturator nerve extracts on total and 16S acetylcholinesterase (AChE) activity were studied in endplate regions of denervated anterior gracilis muscles maintained in organ culture for 48 hr. The decrease of total AChE activity in cultured muscles was similar to that observed in denervated muscles in vivo. This decrease in activity was partly prevented by addition of either 100 or 200 microliters nerve extract (2.7 mg/ml protein) to the nutrient medium. Nerve extract treatment also decreased the release of AChE activity from the muscle into the bathing medium. Conversely, rat serum (20 microliters: 90 mg/ml protein) had no effect on total AChE activity in muscle endplates, nor on release of the enzyme by the muscle. The 16S form of AChE was confined to motor endplate muscle regions and its activity was drastically decreased by denervation in both organ culture and in vivo preparations in a comparable manner. Nerve-extract supplemented cultures contained a significantly (p < 0.001) larger amount of endplate 16S AChE activity (140--145%) than the corresponding controls (100%). Our results suggest that some nerve soluble substance, other than serum contaminants or 16S AChE itself, affects the maintenance of 16S AChE at the neuromuscular junction.

Experience of changing pyrogenicity in albumin solutions.

A number of production lots of an intravenous preparation of 4.3 g/dl of human albumin failed to EP pyrogen test when tested in three rabbits. Several lots were reprocessed and subsequently proved satisfactory but over thirty lots were placed in storage and monitored by both the EP test and by using the LAL test over a period up to 650 days. The majority of the lots showed a steady fall in the pyrogen response by both tests. These lots were removed from storage and reprocessed by sterile filtration and dispensing into new glass containers. Further examination of the product showed it to be satisfactory in all regards. The product of several lots was used in treatment episodes involving the use of cell separators and shown to be satisfactory in clinical use. The remaining lots were also issues for general clinical use without incident.

Calibration of the limulus test for blood products.

The Protein Fractionation Centre has been using the Limulus Amoebocyte Lysate test for detecting the presence of pyrogens for some five years. All tests on final plasma products have been carried out in parallel with the EP rabbit pyrogen test. Good correlation between both tests has been obtained with products such as distilled water and crystalloid solutions. However, it is more difficult to correlate results when dealing with plasma proteins. Each product appears to react differently. Some normally pass the limulus test undiluted while others, such as immune globulins and albuminoid products, have to be diluted to various concentrations before they pass. A preliminary statistical analysis has been carried out on the results of pyrogen testing on Stable Plasma Protein Solution, firstly to ascertain that a significant correlation does exist between the LAL test and the rabbit test, and secondly to try to obtain "cut off" values for the limulus test which compare well with the official EP temperature rises in the rabbit test.