Kerry J. Halupka
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Kerry J. Halupka has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
- Mohit Naresh Shivdasani
School of Biomedical Engineering, University of New South Wales, Sydney, Australia
4 shared publications Anthony N Burkitt
NeuroEngineering Laboratory, Department of Electrical & Electronic Engineering, The University of Melbourne, VIC 3010, Australia
3 shared publications- Benjamin Seth Cazzolato
The University of Adelaide
3 shared publications David B Grayden
NeuroEngineering Laboratory, Department of Electrical & Electronic Engineering, The University of Melbourne, VIC 3010, Australia
3 shared publications- Hamish Meffin
Department of Biomedical Engineering, The University of Melbourne, Parkville, Victoria, Australia
3 shared publications - Steven. D. Wiederman
School of Biomedicine, The University of Adelaide, Adelaide, Australia
3 shared publications
Latest Publications
- 7 October 2019
- Published by ASME International in Journal of Biomechanical Engineering
- Vol. 142 (2)
Using Contrast Motion to Generate Patient-Specific Blood Flow Simulations During Invasive Coronary Angiography
Virtual fractional flow reserve (vFFR) is an emerging technology employing patient-specific computational fluid dynamics (CFD) simulations to infer the hemodynamic significance of a coronary stenosis. Patient-specific boundary conditions are an important aspect of this approach and while most efforts make use of lumped parameter models to capture key phenomena, they lack the ability to specify the associated parameters on a patient-specific basis. When applying vFFR in a catheter laboratory setting using X-ray angiograms as the basis for creating the simulations, there is some indirect functional information available through the observation of the radio-opaque contrast agent motion. In this work, we present a novel method for tuning the lumped parameter arterial resistances (commonly incorporated in such simulations), based on simulating the physics of the contrast motion and comparing the observed and simulated arrival times of the contrast front at key points within a coronary tree. We present proof of principle results on a synthetically generated coronary tree comprised of multiple segments, demonstrating that the method can successfully optimize the arterial resistances to reconstruct the underlying velocity and pressure fields, providing a potential new means to improve the patient specificity of simulation-based technologies in this area.
- 7 October 2019
- Published by ASME International in Journal of Biomechanical Engineering
- Vol. 142 (2)
Using Contrast Motion to Generate Patient-Specific Blood Flow Simulations During Invasive Coronary Angiography
Virtual fractional flow reserve (vFFR) is an emerging technology employing patient-specific computational fluid dynamics (CFD) simulations to infer the hemodynamic significance of a coronary stenosis. Patient-specific boundary conditions are an important aspect of this approach and while most efforts make use of lumped parameter models to capture key phenomena, they lack the ability to specify the associated parameters on a patient-specific basis. When applying vFFR in a catheter laboratory setting using X-ray angiograms as the basis for creating the simulations, there is some indirect functional information available through the observation of the radio-opaque contrast agent motion. In this work, we present a novel method for tuning the lumped parameter arterial resistances (commonly incorporated in such simulations), based on simulating the physics of the contrast motion and comparing the observed and simulated arrival times of the contrast front at key points within a coronary tree. We present proof of principle results on a synthetically generated coronary tree comprised of multiple segments, demonstrating that the method can successfully optimize the arterial resistances to reconstruct the underlying velocity and pressure fields, providing a potential new means to improve the patient specificity of simulation-based technologies in this area.
- Open Access
- 13 November 2018
- Published by Optica Publishing Group in Biomedical Optics Express
- Vol. 9 (12) , 6205-6221
Retinal optical coherence tomography image enhancement via deep learning
Optical coherence tomography (OCT) images of the retina are a powerful tool for diagnosing and monitoring eye disease. However, they are plagued by speckle noise, which reduces image quality and reliability of assessment. This paper introduces a novel speckle reduction method inspired by the recent successes of deep learning in medical imaging. We present two versions of the network to reflect the needs and preferences of different end-users. Specifically, we train a convolution neural network to denoise cross-sections from OCT volumes of healthy eyes using either (1) mean-squared error, or (2) a generative adversarial network (GAN) with Wasserstein distance and perceptual similarity. We then interrogate the success of both methods with extensive quantitative and qualitative metrics on cross-sections from both healthy and glaucomatous eyes. The results show that the former approach provides state-of-the-art improvement in quantitative metrics such as PSNR and SSIM, and aids layer segmentation. However, the latter approach, which puts more weight on visual perception, outperformed for qualitative comparisons based on accuracy, clarity, and personal preference. Overall, our results demonstrate the effectiveness and efficiency of a deep learning approach to denoising OCT images, while maintaining subtle details in the images.
- 13 November 2018
- Published by Optica Publishing Group in Biomedical Optics Express
- Vol. 9 (12) , 6205-6221
Open AccessRetinal optical coherence tomography image enhancement via deep learning
Optical coherence tomography (OCT) images of the retina are a powerful tool for diagnosing and monitoring eye disease. However, they are plagued by speckle noise, which reduces image quality and reliability of assessment. This paper introduces a novel speckle reduction method inspired by the recent successes of deep learning in medical imaging. We present two versions of the network to reflect the needs and preferences of different end-users. Specifically, we train a convolution neural network to denoise cross-sections from OCT volumes of healthy eyes using either (1) mean-squared error, or (2) a generative adversarial network (GAN) with Wasserstein distance and perceptual similarity. We then interrogate the success of both methods with extensive quantitative and qualitative metrics on cross-sections from both healthy and glaucomatous eyes. The results show that the former approach provides state-of-the-art improvement in quantitative metrics such as PSNR and SSIM, and aids layer segmentation. However, the latter approach, which puts more weight on visual perception, outperformed for qualitative comparisons based on accuracy, clarity, and personal preference. Overall, our results demonstrate the effectiveness and efficiency of a deep learning approach to denoising OCT images, while maintaining subtle details in the images.
- Open Access
- 1 July 2017
- Published by Association for Research in Vision and Ophthalmology (ARVO) in Investigative Ophthalmology & Visual Science
- Vol. 58 (9) , 3770-3784
Neural Responses to Multielectrode Stimulation of Healthy and Degenerate Retina
Purpose: Simultaneous stimulation of multiple retinal electrodes in normally sighted animals shows promise in improving the resolution of retinal prostheses. However, the effects of simultaneous stimulation on degenerate retinae remain unknown. Therefore, we investigated the characteristics of cortical responses to multielectrode stimulation of the degenerate retina. Methods: Four adult cats were bilaterally implanted with retinal electrode arrays in the suprachoroidal space after unilateral adenosine triphosphate (ATP)–induced retinal photoreceptor degeneration. Functional and structural changes were characterized by using electroretinogram a-wave amplitude and optical coherence tomography. Multiunit activity was recorded from both hemispheres of the visual cortex. Responses to single- and multielectrode stimulation of the ATP-injected and fellow control eyes were characterized and compared. Results: The retinae of ATP-injected eyes displayed structural and functional changes consistent with mid- to late-stage photoreceptor degeneration and remodeling. Responses to multielectrode stimulation of the ATP-injected eyes exhibited shortened latencies, lower saturated spike counts, and higher thresholds, compared to stimulation of the fellow control eyes. Electrical receptive field sizes were significantly larger in the ATP-injected eye than in the control eye, and positively correlated with the extent of degeneration. Conclusions: Significant differences exist between cortical responses to stimulation of healthy and degenerate retinae. Our results highlight the importance of using a retinal degeneration model when evaluating the efficacy of novel stimulation paradigms.
- 1 July 2017
- Published by Association for Research in Vision and Ophthalmology (ARVO) in Investigative Ophthalmology & Visual Science
- Vol. 58 (9) , 3770-3784
Open AccessNeural Responses to Multielectrode Stimulation of Healthy and Degenerate Retina
Purpose: Simultaneous stimulation of multiple retinal electrodes in normally sighted animals shows promise in improving the resolution of retinal prostheses. However, the effects of simultaneous stimulation on degenerate retinae remain unknown. Therefore, we investigated the characteristics of cortical responses to multielectrode stimulation of the degenerate retina. Methods: Four adult cats were bilaterally implanted with retinal electrode arrays in the suprachoroidal space after unilateral adenosine triphosphate (ATP)–induced retinal photoreceptor degeneration. Functional and structural changes were characterized by using electroretinogram a-wave amplitude and optical coherence tomography. Multiunit activity was recorded from both hemispheres of the visual cortex. Responses to single- and multielectrode stimulation of the ATP-injected and fellow control eyes were characterized and compared. Results: The retinae of ATP-injected eyes displayed structural and functional changes consistent with mid- to late-stage photoreceptor degeneration and remodeling. Responses to multielectrode stimulation of the ATP-injected eyes exhibited shortened latencies, lower saturated spike counts, and higher thresholds, compared to stimulation of the fellow control eyes. Electrical receptive field sizes were significantly larger in the ATP-injected eye than in the control eye, and positively correlated with the extent of degeneration. Conclusions: Significant differences exist between cortical responses to stimulation of healthy and degenerate retinae. Our results highlight the importance of using a retinal degeneration model when evaluating the efficacy of novel stimulation paradigms.
- 30 November 2016
- Published by IOP Publishing in Journal of Neural Engineering
- Vol. 14 (1) , 016006
Prediction of cortical responses to simultaneous electrical stimulation of the retina
Objective. Simultaneous electrical stimulation of multiple electrodes has shown promise in diversifying the responses that can be elicited by retinal prostheses compared to interleaved single electrode stimulation. However, the effects of interactions between electrodes are not well understood and clinical trials with simultaneous stimulation have produced inconsistent results. We investigated the effects of multiple electrode stimulation of the retina by developing a model of cortical responses to retinal stimulation. Approach. Electrical stimuli consisting of temporally sparse, biphasic current pulses, with amplitudes sampled from a bi-dimensional Gaussian distribution, were simultaneously delivered to the retina across a 42-channel electrode array implanted in the suprachoroidal space of anesthetized cats. Visual cortex activity was recorded using penetrating microelectrode arrays. These data were used to identify a linear–nonlinear model of cortical responses to retinal stimulation. The ability of the model to generalize was tested by predicting responses to non-white patterned stimuli. Main results. The model accurately predicted two cortical activity measures: multi-unit neural responses and evoked potential responses to white noise stimuli. The model also provides information about electrical receptive fields, including the relative effects of each stimulating electrode on every recording site. Significance. We have demonstrated a simple model that accurately describes cortical responses to simultaneous stimulation of a suprachoroidal retinal prosthesis. Overall, our results demonstrate that cortical responses to simultaneous multi-electrode stimulation of the retina are repeatable and predictable, and that interactions between electrodes during simultaneous stimulation are predominantly linear. The model shows promise for determining optimal stimulation paradigms for exploiting interactions between electrodes to shape neural activity, thereby improving outcomes for patients with retinal prostheses.
- 30 November 2016
- Published by IOP Publishing in Journal of Neural Engineering
- Vol. 14 (1) , 016006
Prediction of cortical responses to simultaneous electrical stimulation of the retina
Objective. Simultaneous electrical stimulation of multiple electrodes has shown promise in diversifying the responses that can be elicited by retinal prostheses compared to interleaved single electrode stimulation. However, the effects of interactions between electrodes are not well understood and clinical trials with simultaneous stimulation have produced inconsistent results. We investigated the effects of multiple electrode stimulation of the retina by developing a model of cortical responses to retinal stimulation. Approach. Electrical stimuli consisting of temporally sparse, biphasic current pulses, with amplitudes sampled from a bi-dimensional Gaussian distribution, were simultaneously delivered to the retina across a 42-channel electrode array implanted in the suprachoroidal space of anesthetized cats. Visual cortex activity was recorded using penetrating microelectrode arrays. These data were used to identify a linear–nonlinear model of cortical responses to retinal stimulation. The ability of the model to generalize was tested by predicting responses to non-white patterned stimuli. Main results. The model accurately predicted two cortical activity measures: multi-unit neural responses and evoked potential responses to white noise stimuli. The model also provides information about electrical receptive fields, including the relative effects of each stimulating electrode on every recording site. Significance. We have demonstrated a simple model that accurately describes cortical responses to simultaneous stimulation of a suprachoroidal retinal prosthesis. Overall, our results demonstrate that cortical responses to simultaneous multi-electrode stimulation of the retina are repeatable and predictable, and that interactions between electrodes during simultaneous stimulation are predominantly linear. The model shows promise for determining optimal stimulation paradigms for exploiting interactions between electrodes to shape neural activity, thereby improving outcomes for patients with retinal prostheses.
- 30 March 2016
- Published by IOP Publishing in Journal of Neural Engineering
- Vol. 13 (3) , 036003
Spectral distribution of local field potential responses to electrical stimulation of the retina
Objective. Different frequency bands of the local field potential (LFP) have been shown to reflect neuronal activity occurring at varying cortical scales. As such, recordings of the LFP may offer a novel way to test the efficacy of neural prostheses and allow improvement of stimulation strategies via neural feedback. Here we use LFP measurements from visual cortex to characterize neural responses to electrical stimulation of the retina. We aim to show that the LFP is a viable signal that contains sufficient information to optimize the performance of sensory neural prostheses. Approach. Clinically relevant electrode arrays were implanted in the suprachoroidal space of one eye in four felines. LFPs were simultaneously recorded in response to stimulation of individual electrodes using penetrating microelectrode arrays from the visual cortex. The frequency response of each electrode was extracted using multi-taper spectral analysis and the uniqueness of the responses was determined via a linear decoder. Main results. We found that cortical LFPs are reliably modulated by electrical stimulation of the retina and that the responses are spatially localized. We further characterized the spectral distribution of responses, with maximum information being contained in the low and high gamma bands. Finally, we found that LFP responses are unique to a large range of stimulus parameters (∼40) with a maximum conveyable information rate of 6.1 bits. Significance. These results show that the LFP can be used to validate responses to electrical stimulation of the retina and we provide the first steps towards using these responses to provide more efficacious stimulation strategies.
- 30 March 2016
- Published by IOP Publishing in Journal of Neural Engineering
- Vol. 13 (3) , 036003
Spectral distribution of local field potential responses to electrical stimulation of the retina
Objective. Different frequency bands of the local field potential (LFP) have been shown to reflect neuronal activity occurring at varying cortical scales. As such, recordings of the LFP may offer a novel way to test the efficacy of neural prostheses and allow improvement of stimulation strategies via neural feedback. Here we use LFP measurements from visual cortex to characterize neural responses to electrical stimulation of the retina. We aim to show that the LFP is a viable signal that contains sufficient information to optimize the performance of sensory neural prostheses. Approach. Clinically relevant electrode arrays were implanted in the suprachoroidal space of one eye in four felines. LFPs were simultaneously recorded in response to stimulation of individual electrodes using penetrating microelectrode arrays from the visual cortex. The frequency response of each electrode was extracted using multi-taper spectral analysis and the uniqueness of the responses was determined via a linear decoder. Main results. We found that cortical LFPs are reliably modulated by electrical stimulation of the retina and that the responses are spatially localized. We further characterized the spectral distribution of responses, with maximum information being contained in the low and high gamma bands. Finally, we found that LFP responses are unique to a large range of stimulus parameters (∼40) with a maximum conveyable information rate of 6.1 bits. Significance. These results show that the LFP can be used to validate responses to electrical stimulation of the retina and we provide the first steps towards using these responses to provide more efficacious stimulation strategies.
- Open Access
- 21 July 2014
- Published by Springer Nature in BMC Neuroscience
- Vol. 15 (S1) , P95
A linear-nonlinear model accurately predicts cortical responses to simultaneous electrical stimulation with a retinal implant
- 21 July 2014
- Published by Springer Nature in BMC Neuroscience
- Vol. 15 (S1) , P95
Open AccessA linear-nonlinear model accurately predicts cortical responses to simultaneous electrical stimulation with a retinal implant
- 1 September 2013
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 15224880,p. 4098-4102
Bio-inspired feature extraction and enhancement of targets moving against visual clutter during closed loop pursuit
We developed a biologically inspired model for detection and pursuit of small targets against complex backgrounds and tested it in a closed-loop flight arena. A winner-takes-all network of local feature detectors based on insect Small Target Motion Detector (STMD) neurons was used to direct the gaze of a camera mimicking the viewpoint of the pursuer in a series of small steps (saccades) whilst fixating the background. The output of a direction-selective network of 2 nd order local motion correlators was then used to enhance the relative salience of features in the direction of travel of the winning feature. The combination of saccadic fixation and robust target-ground discrimination provided by the STMD front-end, with an attention-like 2 nd order salience enhancement provided very reliable capture of tiny targets even in visually challenging scenarios.
- 1 September 2013
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 15224880,p. 4098-4102
Bio-inspired feature extraction and enhancement of targets moving against visual clutter during closed loop pursuit
We developed a biologically inspired model for detection and pursuit of small targets against complex backgrounds and tested it in a closed-loop flight arena. A winner-takes-all network of local feature detectors based on insect Small Target Motion Detector (STMD) neurons was used to direct the gaze of a camera mimicking the viewpoint of the pursuer in a series of small steps (saccades) whilst fixating the background. The output of a direction-selective network of 2 nd order local motion correlators was then used to enhance the relative salience of features in the direction of travel of the winning feature. The combination of saccadic fixation and robust target-ground discrimination provided by the STMD front-end, with an attention-like 2 nd order salience enhancement provided very reliable capture of tiny targets even in visually challenging scenarios.
- Open Access
- 1 January 2013
- Published by Frontiers Media SA in Frontiers in Physiology
- Vol. 4
Local facilitation improves success in closed loop simulations of insect small target pursuit
Detecting and intercepting a small target is a computationally challenging task, but one solved elegantly by the small brain of flying insects, which have evolved several strategies for pursuit of either prey or conspecifics [1, 2]. Male houseflies track other flies, constantly updating their heading towards the target, based on an error angle between target and pursuer [3, 4], thus mimicking movements of the target on a spiralling flight path that ends in capture. By contrast, predatory dragonflies use an interception strategy, steering to minimise movement of the prey's image on the retina, resulting in a collision course [5]. To evaluate these two strategies for target pursuit against visual clutter, we developed a closed-loop model inspired by insect STMD (Small Target Motion Detector) neurons and pursuit behaviour, set in natural scenery. This allows quantification of the merits of alternative pursuit strategies and of key stages of visual processing that cannot yet be obtained from in-vivo analysis. We found that the velocity tuning of STMD neurons [6, 7] imposed a lower bound on the discriminable velocity of targets. However the intercept method enabled successful pursuits even when the target was moving significantly faster than pursuer. The preliminary stage of the detection algorithm is readily distracted by false positives generated by complex backgrounds. However, addition of a facilitation mechanism, inspired directly by recent physiological analysis of dragonfly STMD neurons [8, 9] prevents such breakthrough responses by amplifying the weak signal of tiny targets moving on long trajectories. With this additional 2nd order salience-enhancing algorithm, we saw a significant improvement in successful target interception, from just over 50% to almost 80% (Chi square=5.87, p=0.015, n=44, Fig. 1).
- 1 January 2013
- Published by Frontiers Media SA in Frontiers in Physiology
- Vol. 4
Open AccessLocal facilitation improves success in closed loop simulations of insect small target pursuit
Detecting and intercepting a small target is a computationally challenging task, but one solved elegantly by the small brain of flying insects, which have evolved several strategies for pursuit of either prey or conspecifics [1, 2]. Male houseflies track other flies, constantly updating their heading towards the target, based on an error angle between target and pursuer [3, 4], thus mimicking movements of the target on a spiralling flight path that ends in capture. By contrast, predatory dragonflies use an interception strategy, steering to minimise movement of the prey's image on the retina, resulting in a collision course [5]. To evaluate these two strategies for target pursuit against visual clutter, we developed a closed-loop model inspired by insect STMD (Small Target Motion Detector) neurons and pursuit behaviour, set in natural scenery. This allows quantification of the merits of alternative pursuit strategies and of key stages of visual processing that cannot yet be obtained from in-vivo analysis. We found that the velocity tuning of STMD neurons [6, 7] imposed a lower bound on the discriminable velocity of targets. However the intercept method enabled successful pursuits even when the target was moving significantly faster than pursuer. The preliminary stage of the detection algorithm is readily distracted by false positives generated by complex backgrounds. However, addition of a facilitation mechanism, inspired directly by recent physiological analysis of dragonfly STMD neurons [8, 9] prevents such breakthrough responses by amplifying the weak signal of tiny targets moving on long trajectories. With this additional 2nd order salience-enhancing algorithm, we saw a significant improvement in successful target interception, from just over 50% to almost 80% (Chi square=5.87, p=0.015, n=44, Fig. 1).
- 1 December 2011
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 143-148
Discrete implementation of biologically inspired image processing for target detection
In nature, systems which visually process the world around them, in computationally efficient manners, have evolved over millions of years. The brain of an insect, which is smaller than a grain of rice, and with less than a million neurons, can effectively engage in computationally challenging tasks. For example, visually detecting and discriminating small moving objects, which are embedded within a complex optical flow pattern (induced by ego-motion). This task has yet to be perfected by current image processing techniques, though recent research is taking inspiration from nature to do so, in creating biologically inspired models of insect vision. This paper presents the progress made on our previous computational model based on electrophysiological data of a class of cells called Small Target Motion Detection neurons (STMDs). This model was based in the continuous temporal domain with constraints imposed on the inputs to the model. Modifications to the model include re-implementation in the discrete domain, the addition of a more physiologically accurate log-normal filter, the inclusion of a Reichardt Correlator and the creation of the highly controllable virtual world as a front end to the model. Model outputs show that the target detecting characteristics of the previous continuous model are maintained, though in a form which is directly applicable to hardware implementation.
- 1 December 2011
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 143-148
Discrete implementation of biologically inspired image processing for target detection
In nature, systems which visually process the world around them, in computationally efficient manners, have evolved over millions of years. The brain of an insect, which is smaller than a grain of rice, and with less than a million neurons, can effectively engage in computationally challenging tasks. For example, visually detecting and discriminating small moving objects, which are embedded within a complex optical flow pattern (induced by ego-motion). This task has yet to be perfected by current image processing techniques, though recent research is taking inspiration from nature to do so, in creating biologically inspired models of insect vision. This paper presents the progress made on our previous computational model based on electrophysiological data of a class of cells called Small Target Motion Detection neurons (STMDs). This model was based in the continuous temporal domain with constraints imposed on the inputs to the model. Modifications to the model include re-implementation in the discrete domain, the addition of a more physiologically accurate log-normal filter, the inclusion of a Reichardt Correlator and the creation of the highly controllable virtual world as a front end to the model. Model outputs show that the target detecting characteristics of the previous continuous model are maintained, though in a form which is directly applicable to hardware implementation.