Expert-level automated malaria diagnosis on routine blood films with deep neural networks.

Over 200 million malaria cases globally lead to half a million deaths annually. Accurate malaria diagnosis remains a challenge. Automated imaging processing approaches to analyze Thick Blood Films (TBF) could provide scalable solutions, for urban healthcare providers in the holoendemic malaria sub-Saharan region. Although several approaches have been attempted to identify malaria parasites in TBF, none have achieved negative and positive predictive performance suitable for clinical use in the west sub-Saharan region. While malaria parasite object detection remains an intermediary step in achieving automatic patient diagnosis, training state-of-the-art deep-learning object detectors requires the human-expert labor-intensive process of labeling a large dataset of digitized TBF. To overcome these challenges and to achieve a clinically usable system, we show a novel approach. It leverages routine clinical-microscopy labels from our quality-controlled malaria clinics, to train a Deep Malaria Convolutional Neural Network classifier (DeepMCNN) for automated malaria diagnosis. Our system also provides total Malaria Parasite (MP) and White Blood Cell (WBC) counts allowing parasitemia estimation in MP/µL, as recommended by the WHO. Prospective validation of the DeepMCNN achieves sensitivity/specificity of 0.92/0.90 against expert-level malaria diagnosis. Our approach PPV/NPV performance is of 0.92/0.90, which is clinically usable in our holoendemic settings in the densely populated metropolis of Ibadan. It is located within the most populous African country (Nigeria) and with one of the largest burdens of Plasmodium falciparum malaria. Our openly available method is of importance for strategies aimed to scale malaria diagnosis in urban regions where daily assessment of thousands of specimens is required.

Coding source localization through inter-spike delay: modelling a cluster of Pacinian Corpuscles using time-division multiplexing approach.

The Pacinian Corpuscle (PC) is the most sensitive mechanoreceptor in the human body found in clusters of two or three. We extended our previous model of an isolated-PC to a cluster-PC focussing on relative spike delay and displacement threshold for understanding how the stimulus location is coded. In our model, two PCs with Gaussian overlapping receptive fields are arranged beneath the skin model. For a spatiotemporal stimulus (vibration), the model response is proposed to be a time-division multiplexing of responses from two PCs within the cluster. While the spike rate characteristics and pole-zero plot of cluster-PC model show similarities with the isolated-PC model, the frequency response shows ripples after 1 kHz as the distance (d) between the PCs increases. The stimulus location [Formula: see text] and d influence the relative spike delay and the displacement threshold, but not the spike rate. The novel contributions from our model include prediction of (i) relative spike delay for various d, stimulus frequency (f), and ψ, (ii) spike rate characteristics for varying f, and (iii) displacement threshold curve as a function of frequency for various d. Although the physiological validation of the novel predictions is impractical, we have validated the relative spike delay and the displacement threshold curves with experimental data in the literature.

In-vivo high resolution AFM topographic imaging of Caenorhabditis elegans reveals previously unreported surface structures of cuticle mutants.

Atomic force microscopy (AFM) is a powerful method for topographic imaging of surfaces with nanometer resolution. AFM offers significant advantages over scanning electron microscopy (SEM) including the acquisition of quantitative 3D-images and biomechanical information. More importantly, for in-vivo biological imaging, AFM does not require sample dehydration/labeling. We show for the first time high-resolution topographical images of the cuticle of the model organism C. elegans under physiological conditions using AFM. C. elegans is used extensively for drug screening and to study pathogen adherence in innate immunity; both applications highly depend on the integrity of the nematode's cuticle. Mutations affecting both drug adsorption and pathogen clearance have been proposed to relate to changes in the cuticle structure, but never visually examined in high resolution. In this study we use AFM to visualize the topography of wild-type adult C. elegans as well as several cuticle collagen mutants and describe previously unseen anatomical differences.

Viscoelastic characterization of the primate finger pad in vivo by microstep indentation and three-dimensional finite element models for tactile sensation studies.

When we touch an object, surface loads imposed on the skin are transmitted to thousands of specialized nerve endings (mechanoreceptors) embedded within the skin. These mechanoreceptors transduce the mechanical signals imposed on them into a neural code of the incident stimuli, enabling us to feel the object. To understand the mechanisms of tactile sensation, it is critical to understand the relationship between the applied surface loads, mechanical state at the mechanoreceptor locations, and transduced neural codes. In this paper, we characterize the bulk viscoelastic properties of the primate finger pad and show its relationship to the dynamic firing rate of SA-1 mechanoreceptors. Two three-dimensional (3D) finite element viscoelastic models, a homogeneous and a multilayer model, of the primate fingertip are developed and calibrated with data from a series of force responses to micro-indentation experiments on primate finger pads. We test these models for validation by simulating indentation with a line load and comparing surface deflection with data in the literature (Srinivasan, 1989, "Surface Deflection of Primate Fingertip Under Line Load," J. Biomech., 22(4), pp. 343-349). We show that a multilayer model with an elastic epidermis and viscoelastic core predicts both the spatial and temporal biomechanical response of the primate finger pad. Finally, to show the utility of the model, ramp and hold indentation with a flat plate is simulated. The multilayer model predicts the strain energy density at a mechanoreceptor location would decay at the same rate as the average dynamic firing rate of SA-1 mechanoreceptors in response to flat plate indentation (previously observed by Srinivasan and LaMotte, 1991 "Encoding of Shape in the Responses of Cutaneous Mechanoreceptors," Information Processing in the Somatosensory System (Wenner-Gren International Symposium Series), O. Franzen and J. Westman, eds., Macmillan Press, London, UK), suggesting that the rate of adaptation of SA-1 mechanoreceptors is governed by the viscoelastic nature of its surrounding tissue.

Rehabilitation Program Integrating Virtual Environment to Improve Orientation and Mobility Skills for People Who Are Blind.

This paper presents the integration of a virtual environment (BlindAid) in an orientation and mobility rehabilitation program as a training aid for people who are blind. BlindAid allows the users to interact with different virtual structures and objects through auditory and haptic feedback. This research explores if and how use of the BlindAid in conjunction with a rehabilitation program can help people who are blind train themselves in familiar and unfamiliar spaces. The study, focused on nine participants who were congenitally, adventitiously, and newly blind, during their orientation and mobility rehabilitation program at the Carroll Center for the Blind (Newton, Massachusetts, USA). The research was implemented using virtual environment (VE) exploration tasks and orientation tasks in virtual environments and real spaces. The methodology encompassed both qualitative and quantitative methods, including interviews, a questionnaire, videotape recording, and user computer logs. The results demonstrated that the BlindAid training gave participants additional time to explore the virtual environment systematically. Secondly, it helped elucidate several issues concerning the potential strengths of the BlindAid system as a training aid for orientation and mobility for both adults and teenagers who are congenitally, adventitiously, and newly blind.

The influence of internal pressure and neuromuscular agents on C. elegans biomechanics: an empirical and multi-compartmental in silico modelling study.

The function of a specific tissue and its biomechanics are interdependent, with pathologies or ageing often being intertwined with structural decline. The biomechanics of Caenorhabditis elegans, a model organism widely used in pharmacological and ageing research, has been established as biomarker for healthy ageing. However, the properties of the constituent tissues, and their contribution to the overall mechanical characteristics of the organism, remain relatively unknown. In this study we investigated the biomechanics of healthy C. elegans cuticle, muscle tissue, and pseudocoelom using a combination of indentation experiments and in silico modelling. We performed stiffness measurements using an atomic force microscope. To approximate the nematode's cylindrical body we used a novel three-compartment nonlinear finite element model, enabling us to analyse of how changes in the elasticity of individual compartments affect the bulk stiffness. We then fine-tuned the parameters of the model to match the simulation force-indentation output to the experimental data. To test the finite element model, we modified distinct compartments experimentally. Our in silico results, in agreement with previous studies, suggest that hyperosmotic shock reduces stiffness by decreasing the internal pressure. Unexpectedly, treatment with the neuromuscular agent aldicarb, traditionally associated with muscle contraction, reduced stiffness by decreasing the internal pressure. Furthermore, our finite element model can offer insights into how drugs, mutations, or processes such as ageing target individual tissues.

Perception of Soft Objects in Virtual Environments Under Conflicting Visual and Haptic Cues.

In virtual/augmented/mixed reality (VR/AR/MR) applications, rendering soft virtual objects using a hand-held haptic device is challenging due to the anatomical restrictions of the hand and the ungrounded nature of the design, which affect the selection of actuators and sensors and hence limit the resolution and range of forces displayed by the device. We developed a cable-driven haptic device for rendering the net forces involved in grasping and squeezing 3D virtual compliant (soft) objects being held between the index finger and thumb only. Using the proposed device, we investigate the perception of soft objects in virtual environments. We show that the range of object stiffness that can be effectively conveyed to a user in virtual environments (VEs) can be significantly expanded by controlling the relationship between the visual and haptic cues. We propose that a single variable, named Apparent Stiffness Difference, can predict the pattern of human stiffness perception under manipulated conflict, which can be used for rendering a range of soft objects in VEs larger than what is achievable by a haptic device alone due to its physical limits.

Mechanical properties measured by atomic force microscopy define health biomarkers in ageing C. elegans.

Genetic and environmental factors are key drivers regulating organismal lifespan but how these impact healthspan is less well understood. Techniques capturing biomechanical properties of tissues on a nano-scale level are providing new insights into disease mechanisms. Here, we apply Atomic Force Microscopy (AFM) to quantitatively measure the change in biomechanical properties associated with ageing Caenorhabditis elegans in addition to capturing high-resolution topographical images of cuticle senescence. We show that distinct dietary restriction regimes and genetic pathways that increase lifespan lead to radically different healthspan outcomes. Hence, our data support the view that prolonged lifespan does not always coincide with extended healthspan. Importantly, we identify the insulin signalling pathway in C. elegans and interventions altering bacterial physiology as increasing both lifespan and healthspan. Overall, AFM provides a highly sensitive technique to measure organismal biomechanical fitness and delivers an approach to screen for health-improving conditions, an essential step towards healthy ageing.

Data-driven malaria prevalence prediction in large densely populated urban holoendemic sub-Saharan West Africa.

Over 200 million malaria cases globally lead to half-million deaths annually. The development of malaria prevalence prediction systems to support malaria care pathways has been hindered by lack of data, a tendency towards universal "monolithic" models (one-size-fits-all-regions) and a focus on long lead time predictions. Current systems do not provide short-term local predictions at an accuracy suitable for deployment in clinical practice. Here we show a data-driven approach that reliably produces one-month-ahead prevalence prediction within a densely populated all-year-round malaria metropolis of over 3.5 million inhabitants situated in Nigeria which has one of the largest global burdens of P. falciparum malaria. We estimate one-month-ahead prevalence in a unique 22-years prospective regional dataset of > 9 × 104 participants attending our healthcare services. Our system agrees with both magnitude and direction of the prediction on validation data achieving MAE ≤ 6 × 10-2, MSE ≤ 7 × 10-3, PCC (median 0.63, IQR 0.3) and with more than 80% of estimates within a (+ 0.1 to - 0.05) error-tolerance range which is clinically relevant for decision-support in our holoendemic setting. Our data-driven approach could facilitate healthcare systems to harness their own data to support local malaria care pathways.

Three-dimensional behavioural phenotyping of freely moving C. elegans using quantitative light field microscopy.

Behavioural phenotyping of model organisms is widely used to investigate fundamental aspects of organism biology, from the functioning of the nervous system to the effects of genetic mutations, as well as for screening new drug compounds. However, our capacity to observe and quantify the full range and complexity of behavioural responses is limited by the inability of conventional microscopy techniques to capture volumetric image information at sufficient speed. In this article we describe how combining light field microscopy with computational depth estimation provides a new method for fast, quantitative assessment of 3D posture and movement of the model organism Caenorhabditis elegans (C. elegans). We apply this technique to compare the behaviour of cuticle collagen mutants, finding significant differences in 3D posture and locomotion. We demonstrate the ability of quantitative light field microscopy to provide new fundamental insights into C. elegans locomotion by analysing the 3D postural modes of a freely swimming worm. Finally, we consider relative merits of the method and its broader application for phenotypic imaging of other organisms and for other volumetric bioimaging applications.

Natural Infection of C. elegans by an Oomycete Reveals a New Pathogen-Specific Immune Response.

In its natural habitat, the nematode Caenorhabditis elegans encounters a plethora of other organisms, including many that are pathogenic [1, 2]. The study of interactions between C. elegans and various pathogens has contributed to characterizing key mechanisms of innate immunity [2-4]. However, how C. elegans recognizes different pathogens to mount pathogen-specific immune responses remains still largely unknown [3, 5-8]. Expanding the range of known C. elegans-infecting pathogens and characterizing novel pathogen-specific immune responses are key steps toward answering this question. We report here that the oomycete Myzocytiopsis humicola is a natural pathogen of C. elegans, and we describe its infection strategy. We identify a new host immune response to pathogen exposure that involves induction of members of a previously uncharacterized gene family encoding chitinase-like (CHIL) proteins. We demonstrate that this response is highly specific against M. humicola and antagonizes the infection. We propose that CHIL proteins may diminish the ability of the oomycete to infect by hindering pathogen attachment to the host cuticle. This work expands our knowledge of natural eukaryotic pathogens of C. elegans and introduces a new pathosystem to address how animal hosts recognize and respond to oomycete infections.

The muscle activation method: an approach to impedance control of brain-machine interfaces through a musculoskeletal model of the arm.

Current demonstrations of brain-machine interfaces (BMIs) have shown the potential for controlling neuroprostheses under pure motion control. For interaction with objects, however, pure motion control lacks the information required for versatile manipulation. This paper investigates the idea of applying impedance control in a BMI system. An extraction algorithm incorporating a musculoskeletal arm model was developed for this purpose. The new algorithm, called the muscle activation method (MAM), was tested on cortical recordings from a behaving monkey. The MAM was found to predict motion parameters with as much accuracy as a linear filter. Furthermore, it successfully predicted limb interactions with novel force fields, which is a new and significant capability lacking in other algorithms.

Determining the biomechanics of touch sensation in C. elegans.

The sense of touch is a fundamental mechanism that nearly all organisms use to interact with their surroundings. However, the process of mechanotransduction whereby a mechanical stimulus gives rise to a neuronal response is not well understood. In this paper we present an investigation of the biomechanics of touch using the model organism C. elegans. By developing a custom micromanipulation and force sensing system around a high resolution optical microscope, we measured the spatial deformation of the organism's cuticle and force response to controlled uniaxial indentations. We combined these experimental results with anatomical data to create a multilayer computational biomechanical model of the organism and accurately derive its material properties such as the elastic modulus and poisson's ratio. We demonstrate the utility of this model by combining it with previously published electrophysiological data to provide quantitative insights into different biomechanical states for mechanotransduction, including the first estimate of the sensitivity of an individual mechanoreceptor to an applied stimulus (parameterised as strain energy density). We also interpret empirical behavioural data to estimate the minimum number of mechanoreceptors which must be activated to elicit a behavioural response.

In vivo mechanical behavior of intra-abdominal organs.

For realistic surgical simulation in a virtual environment, in vivo material properties of biological tissues are required for simulating the deformations and the reaction forces from the tool-tissue interactions. In this paper, the in vivo static and dynamic mechanical behavior of the liver and lower esophagus of pigs were presented both in linear and nonlinear regions under compressive and shear indentations. A robotic device was programmed to function as a mechanical stimulator with a 2-mm flat-tipped cylindrical probe attached to its tip. A series of ramp and hold stimuli, as well as sinusoidal indentation stimuli, were delivered to the organs and reaction forces were measured. The conditions for these indentation stimuli were designed such that they were similar to conditions in an operating room. Experiments were also carried out on the organs for ex vivo and in vitro conditions. Results show that the breathing and pulse rate significantly affect the measured force responses of the organs. From the obtained force-displacement relationships, steady-state impedances as well as dynamic impedances of both organs were calculated. The results also show that in vivo steady-state impedance of the lower esophagus is significantly higher than that of the liver. The in vivo steady-state response of the liver, however, exhibits a greater degree of nonlinearity than that of the lower esophagus. The in vivo steady-state response of the lower esophagus in the three orthogonal directions also indicates that the lower esophagus is not significantly anisotropic. The impedance of both organs under sinusoidal indentations (0-5 Hz) are fairly similar each other. Magnitudes of the impedance over the stimulus frequencies are fairly constant. The impedance phase angles decrease over the range of stimulus frequencies applied. Comparison of the measurements obtained from the in vivo, ex vivo, and in vitro experiments shows that the mechanical properties of the biological tissues change significantly after the death of the animal. The tissues generally become stiffer and exhibit greater nonlinearity. The degree of change in their mechanical properties is dependent on the amount of time after the death of the animal. These data can be further utilized in the computing of the material parameters of tissue models for laparoscopic surgery simulators as well as open surgery simulators.

Continuous shared control for stabilizing reaching and grasping with brain-machine interfaces.

Research on brain-machine interfaces (BMI's) is directed toward enabling paralyzed individuals to manipulate their environment through slave robots. Even for able-bodied individuals, using a robot to reach and grasp objects in unstructured environments can be a difficult telemanipulation task. Controlling the slave directly with neural signals instead of a hand-master adds further challenges, such as uncertainty about the intended trajectory coupled with a low update rate for the command signal. To address these challenges, a continuous shared control (CSC) paradigm is introduced for BMI where robot sensors produce reflex-like reactions to augment brain-controlled trajectories. To test the merits of this approach, CSC was implemented on a 3-degree-of-freedom robot with a gripper bearing three co-located range sensors. The robot was commanded to follow eighty-three reach-and-grasp trajectories estimated previously from the outputs of a population of neurons recorded from the brain of a monkey. Five different levels of sensor-based reflexes were tested. Weighting brain commands 70% and sensor commands 30% produced the best task performance, better than brain signals alone by more than seven-fold. Such a marked performance improvement in this test case suggests that some level of machine autonomy will be an important component of successful BMI systems in general.

Investigation of mechanosensation in C. elegans using light field calcium imaging.

We describe a new experimental approach to investigate touch sensation in the model organism C. elegans using light field deconvolution microscopy. By combining fast volumetric image acquisition with controlled indentation of the organism using a high sensitivity force transducer, we are able to simultaneously measure activity in multiple touch receptor neurons expressing the calcium ion indicator GCaMP6s. By varying the applied mechanical stimulus we show how this method can be used to quantify touch sensitivity in C. elegans. We describe some of the challenges of performing light field calcium imaging in moving samples and demonstrate that they can be overcome by simple data processing.

Nano-mechanical single-cell sensing of cell-matrix contacts.

Extracellular protein matrices provide a rigidity interface exhibiting nano-mechanical cues that guide cell growth and proliferation. Cells sense such cues using actin-rich filopodia extensions which encourage favourable cell-matrix contacts to recruit more actin-mediated local forces into forming stable focal adhesions. A challenge remains in identifying and measuring these local cellular forces and in establishing empirical relationships between them, cell adhesion and filopodia formation. Here we investigate such relationships using a micromanipulation system designed to operate at the time scale of focal contact dynamics, with the sample frequency of a force probe being 0.1 ms, and to apply and measure forces at nano-to-micro Newton ranges for individual mammalian cells. We explore correlations between cell biomechanics, cell-matrix attachment forces and the spread areas of adhered cells as well as their relative dependence on filopodia formation using synthetic protein matrices with a proven ability to induce enhanced filopodia numbers in adherent cells. This study offers a basis for engineering exploitable cell-matrix contacts in situ at the nanoscale and single-cell levels.

Vibrotactile sensitivity threshold: nonlinear stochastic mechanotransduction model of the Pacinian Corpuscle.

Based on recent discoveries of stretch and voltage activated ion channels in the receptive area of the Pacinian Corpuscle (PC), this paper describes a two-stage mechanotransduction model of its near threshold Vibrotactile (VT) sensitivity valid over 10 Hz to a few kHz. The model is based on the nonlinear and stochastic behavior of the ion channels represented as dependent charge sources loaded with membrane impedance. It simulates the neural response of the PC considering the morphological and statistical properties of the receptor potential and action potential with the help of an adaptive relaxation pulse frequency modulator. This model also simulates the plateaus and nonmonotonic saturation of spike rate characteristics. The stochastic simulation based on the addition of mechanical and neural noise describes that the VT Sensitivity Threshold (VTST) at higher frequencies is more noise dependent. Above 800 Hz even a SNR = 150 improves the neurophysiological VTST more than 3 dBµ. In that frequency range, an absence of the entrainment threshold and a lower sensitivity index near the absolute threshold make the upper bound of the psychophysical VTST more dependent on the experimental protocol and physical set-up. This model can be extended to simulate the neural response of a group of PCs.

Multiscale layered biomechanical model of the pacinian corpuscle.

This paper describes a multiscale analytical model of the lamellar structure and the biomechanical response of the Pacinian Corpuscle (PC). In order to analyze the contribution of the PC lamellar structure for detecting high-frequency vibrotactile (VT) stimuli covering 10 Hz to a few kHz, the model response is studied against trapezoidal and sinusoidal stimuli. The model identifies a few generalizable features of the lamellar structure which makes it scalable for different sizes of PC with different number of lamellae. The model describes the mechanical signal conditioning of the lamellar structure in terms of a recursive transfer-function, termed as the Compression-Transmittance-Transfer-Function (CTTF). The analytical results show that with the increase of the PC layer index above 15, the PC inner core (IC) relaxes within 1 ms against step compression of the outermost layer. This model also considers the mass of each PC layer to investigate its effect on the biomechanical response of the lamellar structure. The interlamellar spacing and its biomechanical properties along with the model response are validated with experimental data in the literature. The proposed model can be used for simulating a network of PCs considering their diversity for analyzing the high-frequency VT sensitivity of the human skin.

Quantitative ultrasonic methods for characterization of skin lesions in vivo.

Quantitative ultrasonic methods were studied for characterizing skin lesions in vivo using contact dermatitis as an example. The parameters studied include skin thickness, echogenicity, attenuation coefficient slope and parameters related to echo statistics (signal-to-noise ratio and shape parameters of Weibull, K and generalized gamma distributions). Data were collected using a high-frequency ultrasound (US) system (center frequency = 33 MHz). To compensate for depth-dependent diffraction effects, correction curves as a function of the distance between the transducer and the tissue were first empirically obtained. Diffraction-corrected quantitative parameters were then compared between healthy and affected skin of volunteers, who underwent patch testing for allergic and irritant contact dermatitis. A significant increase in skin thickness, decrease in echogenicity of the upper dermis and decrease in attenuation coefficient slope were found at the affected sites compared to those of healthy skin. However, no differences in parameters related to the echo statistics of the mid-dermis were found. These results indicate that a combination of quantitative ultrasonic parameters have the potential for extracting information for characterizing skin conditions.