McDonald, Andrew; Matos, Jose Novo; Silva, Joel; Partington, Catheryn; Lo, Eve J. Y.; Fuentes, Virginia Luis; Barron, Lara; Watson, Penny; Agarwal, Anurag A machine-learning algorithm to grade heart murmurs and stage preclinical myxomatous mitral valve disease in dogs (Journal Article) In: Journal of Veterinary Internal Medicine, 2024. @article{canine_jvim,Background: The presence and intensity of heart murmurs are sensitive indicators of several cardiac diseases in dogs, particularly myxomatous mitral valve disease (MMVD), but accurate interpretation requires substantial clinical expertise. Objectives: Assess if a machine-learning algorithm can be trained to accurately detect and grade heart murmurs in dogs and detect cardiac disease in electronic stethoscope recordings. Animals: Dogs (n = 756) with and without cardiac disease attending referral centers in the United Kingdom. Methods: All dogs received full physical and echocardiographic examinations by a cardiologist to grade any murmurs and identify cardiac disease. A recurrent neural network algorithm, originally trained for heart murmur detection in humans, was fine-tuned on a subset of the dog data to predict the cardiologist's murmur grade from the audio recordings. Results: The algorithm detected murmurs of any grade with a sensitivity of 87.9% (95% confidence interval [CI], 83.8%-92.1%) and a specificity of 81.7% (95% CI, 72.8%-89.0%). The predicted grade exactly matched the cardiologist's grade in 57.0% of recordings (95% CI, 52.8%-61.0%). The algorithm's prediction of loud or thrilling murmurs effectively differentiated between stage B1 and B2 preclinical MMVD (area under the curve [AUC], 0.861; 95% CI, 0.791-0.922), with a sensitivity of 81.4% (95% CI, 68.3%-93.3%) and a specificity of 73.9% (95% CI, 61.5%-84.9%). Conclusion and Clinical Importance: A machine-learning algorithm trained on humans can be successfully adapted to grade heart murmurs in dogs caused by common cardiac diseases, and assist in differentiating preclinical MMVD. The model is a promising tool to enable accurate, low-cost screening in primary care. |
Bassil, KJ.; Agarwal, A.; Celler, BG Unravelling the mechanics of cuff-based blood pressure measurement and the resulting sources of error (Inproceedings) In: Proceedings of the 26th International Congress of Theoretical and Applied Mechanics, pp. 581-2, IUTAM 2024. @inproceedings{Bassil2024,Auscultatory blood pressure measurement is the gold standard against which other non-invasive measurement devices are calibrated. However, this method systematically underestimates systolic and overestimates diastolic blood pressure. The overestimation is explained by known causes of error - primarily the stiffness of the artery and surrounding tissue. Within this context, the underestimation of systolic pressure seems paradoxical. We show that the systolic measurement is affected by changes to the arterial pressure, occurring when the blood pressure cuff is applied. The occlusion of the artery results in low pressure downstream of the cuff. Through varying parameters in a novel experimental rig, we demonstrate that the lower this downstream pressure, the greater the underestimation of the systolic pressure. Our results suggest a need for a new, more accurate measurement protocol. |
McDonald, A.; Agarwal, A.; Williams, B.; Liu, N-C; Ladlow, J. Neural network analysis of pharyngeal sounds can detect obstructive upper respiratory disease in brachycephalic dogs (Journal Article) In: PLOS ONE, vol. 18, iss. 8, no. e0305633, pp. 1–16, 2024. @article{MAW24, |
Bugeat, Benjamin; Karban, Ugur; Agarwal, Anurag; Lesshafft, Lutz; Jordan, Peter Acoustic resolvent analysis of turbulent jets (Journal Article) In: Theoretical and Computational Fluid Dynamics, vol. 38, pp. 687–706, 2024. (BibTeX) @article{bugeat2024acoustic, |
Bassil, KJ.; Agarwal, A.; Celler, BG. An explanation for underestimation of systolic pressure in cuff-based measurement (Inproceedings) In: Journal of Hypertension, pp. e87, 2024. @inproceedings{nokey, |
Karban, U.; Bugeat, B.; Towne, A.; Lesshafft, L.; Agarwal, A.; Jordan, P. An empirical model of noise sources in subsonic jets (Journal Article) In: Journal of Fluid Mechanics, vol. 965, pp. A18, 2023. @article{KBT23, |
Nussbaumer, M.; Agarwal, A. Stethoscope acoustics (Journal Article) In: Journal of Sound and Vibration, vol. 539, no. 117194, pp. 1–18, 2022. @article{NA2022,For over 200 years, stethoscopes have been used to diagnose disease by listening to the sounds from the human body. However, the physics of how stethoscopes work remains poorly understood, mainly because their performance depends not only on the stethoscopes themselves, but on the coupled system that forms when the hand of a clinician presses the device against the chest of a patient. We develop an experimental setup that allows us to characterise the effect of each constituent part on the behaviour of the coupled system. By using a suitably instrumented ‘chest phantom’ we are able to quantify the response of stethoscopes in a repeatable manner and account for the effect of the phantom on the response. We provide a theoretical framework for understanding the acoustics of the coupled stethoscope system and propose a low-order lumped-element model that captures the effects of the key design choices. For example, minimising the air cavity volume inside the stethoscope maximises the response, while a stethoscope’s tubing significantly attenuates the response and introduces distorting standing-wave resonances. Using a diaphragm attenuates the response and shifts the resonances to higher frequencies. However, it also allows the air cavity volume to be minimised, which can offset the attenuation. We dispel several misconceptions, such as that the stiffness of diaphragms leads to an amplification of higher frequencies or that the dimensions of the rim and the mass of the chestpiece play no role in stethoscope performance. We hope that this paper will help to optimise the design of stethoscopes. |
McDonald, A.; Agarwal, A.; Gales, M. J. F. Detection of heart murmurs in phonocardiograms with parallel hidden semi-Markov models (Proceeding) Proceedings of Computing in Cardiology 2022, 2022. (BibTeX) @proceedings{MAG22, |
W. Jiang J. Håkansson, Q. Xue; Elemans, C. P. H. Aerodynamics and motor control of ultrasonic vocalizations for social communication in mice and rats (Journal Article) In: BMC Biology, vol. 20, no. 3, pp. 1–15, 2022. @article{hakansson2022, |
McDonald, A.; Agarwal, A.; Marr, C. Machine intelligence for the detection of equine heart murmurs (Journal Article) In: Equine Veterinary Journal, vol. 54, pp. 15–16, 2022. (BibTeX) @article{MAM22, |
Gregory, A. L.; Agarwal, A.; Lasenby, J. An experimental investigation to model wheezing in lungs (Journal Article) In: Royal Society Open Science, vol. 8, no. 2, pp. 1–20, 2021. @article{Gregory2021,A quarter of the world's population experience wheezing. These sounds have been used for diagnosis since the time of the Ebers Papyrus (ca 1500 BC). We know that wheezing is a result of the oscillations of the airways that make up the lung. However, the physical mechanisms for the onset of wheezing remain poorly understood, and we do not have a quantitative model to predict when wheezing occurs. We address these issues in this paper. We model the airways of the lungs by a modified Starling resistor in which airflow is driven through thin, stretched elastic tubes. By completing systematic experiments, we find a generalized ‘tube law’ that describes how the cross-sectional area of the tubes change in response to the transmural pressure difference across them. We find the necessary conditions for the onset of oscillations that represent wheezing and propose a flutter-like instability model for it about a heavily deformed state of the tube. Our findings allow for a predictive tool for wheezing in lungs, which could lead to better diagnosis and treatment of lung diseases. |
Anurag Agarwal Martin Thoenes, David Grundmann Narrative review of the role of artificial intelligence to improve aortic valve disease management (Journal Article) In: Journal of Thoracic Disease, vol. 13, no. 1, pp. 396-404, 2021, ISSN: 2077-6624. @article{Thoenes2021,Valvular heart disease (VHD) is a chronic progressive condition with an increasing prevalence in the Western world due to aging populations. VHD is often diagnosed at a late stage when patients are symptomatic and the outcomes of therapy, including valve replacement, may be sub-optimal due the development of secondary complications, including left ventricular (LV) dysfunction. The clinical application of artificial intelligence (AI), including machine learning (ML), has promise in supporting not only early and more timely diagnosis, but also hastening patient referral and ensuring optimal treatment of VHD. As physician auscultation lacks accuracy in diagnosis of significant VHD, computer-aided auscultation (CAA) with the help of a commercially available digital stethoscopes improves the detection and classification of heart murmurs. Although used little in current clinical practice, CAA can screen large populations at low cost with high accuracy for VHD and faciliate appropriate patient referral. Echocardiography remains the next step in assessment and planning management and AI is delivering major changes in speeding training, improving image quality by pattern recognition and image sorting, as well as automated measurement of multiple variables, thereby improving accuracy. Furthermore, AI then has the potential to hasten patient disposal, by automated alerts for red-flag findings, as well as decision support in dealing with results. In management, there is great potential in ML-enabled tools to support comprehensive disease monitoring and individualized treatment decisions. Using data from multiple sources, including demographic and clinical risk data to image variables and electronic reports from electronic medical records, specific patient phenotypes may be identified that are associated with greater risk or modeled to the estimate trajectory of VHD progression. Finally, AI algorithms are of proven value in planning intervention, facilitating transcatheter valve replacement by automated measurements of anatomical dimensions derived from imaging data to improve valve selection, valve size and method of delivery. |
Karban, U.; Bugeat, B.; Martini, E.; Towne, A.; Cavalieri, A. V. G.; Lesshafft, L.; Agarwal, A.; Jordan, P.; Colonius, T. Ambiguity in mean-flow-based linear analysis (Journal Article) In: Journal of Fluid Mechanics, vol. 900, no. R5, 2020. @article{Karban2020,Linearisation of the Navier–Stokes equations about the mean of a turbulent flow forms the foundation of popular models for energy amplification and coherent structures, including resolvent analysis. While the Navier–Stokes equations can be equivalently written using many different sets of dependent variables, we show that the properties of the linear operator obtained via linearisation about the mean depend on the variables in which the equations are written prior to linearisation, and can be modified under nonlinear transformation of variables. For example, we show that using primitive and conservative variables leads to differences in the singular values and modes of the resolvent operator for turbulent jets, and that the differences become more severe as variable-density effects increase. This lack of uniqueness of mean-flow-based linear analysis provides new opportunities for optimising models by specific choice of variables while also highlighting the importance of carefully accounting for the nonlinear terms that act as a forcing on the resolvent operator. |
Nussbaumer, M.; Martinez, L. T.; Agarwal, A. A theory for stethoscope acoustics (Inproceedings) In: Proceedings of the 23rd international congress on acoustics, pp. 4991–4998, 2019, (Best Student Paper Award). @inproceedings{NTA19,Maximising the signal to noise ratio while considering ergonomics and aesthetics is the key design challenge for modern stethoscopes. In order to optimise the design, there is a need for a well-validated model for the transfer function from a source within the chest to the output signal obtained. Most variants of the stethoscope are air- coupled sensors. Here we propose a new theory for the acoustics of this type of sensor, which takes into account the coupling between the sensor and the human chest. We have conducted rigorous experiments to characterise the transfer function of the chest-stethoscope system and have investigated the effects of key design parameters. Our data confirms traditional findings on the effects of bell geometry and diaphragm usage, but also highlights the importance of the coupling between the sensor and the chest, and reveals features of the transfer function that are not captured by existing models. Our model employs a transmission matrix formulation and discretises the system into lumped element components. It can be used to inform design choices for acoustic, electronic and dual-mode stethoscopes, opening up the possibility of an optimum design that maximises the signal to noise ratio for a desired application. |
Phillips, Sam; Agarwal, Anurag; Jordan, Peter The annoying dripping tap (Journal Article) In: Physics Today, vol. 71, no. 12, pp. 70-71, 2018. @article{PAJPh18,Whether produced by a leaky tap dripping into a sink or by raindrops falling onto a lake, the distinctive and at times irritating plink of a small drop falling into a larger body of liquid has been the subject of scientific curiosity for more than a century. In 1989 Hugh Pumphrey and colleagues determined that the source of the sound is a pulsating air bubble trapped under the surface of the water during the drop impact. However, until recently it was not clear just how the underwater bubble produces the plink heard above water by the human ear. In this Quick Study we describe a project that took advantage of devices unavailable to Pumphrey to better understand how the bubble generates its characteristic sound. |
Vincent Jaunet Peter Jordan, Aaron Towne Jet–flap interaction tones (Journal Article) In: Journal of Fluid Mechanics, vol. 853, pp. Pages, 2018. @article{JVT18,Motivated by the problem of jet–flap interaction noise, we study the tonal dynamics that occurs when an isothermal turbulent jet grazes a sharp edge. We perform hydrodynamic and acoustic pressure measurements to characterise the tones as a function of Mach number and streamwise edge position. The observed distribution of spectral peaks cannot be explained using the usual edge-tone model, in which resonance is underpinned by coupling between downstream-travelling Kelvin–Helmholtz wavepackets and upstream-travelling sound waves. We show, rather, that the strongest tones are due to coupling between Kelvin–Helmholtz wavepackets and a family of trapped, upstream-travelling acoustic modes in the potential core, recently studied by Towne et al. (J. Fluid Mech. vol. 825, 2017) and Schmidt et al. (J. Fluid Mech. vol. 825, 2017). We also study the band-limited nature of the resonance, showing the high-frequency cutoff to be due to the frequency dependence of the upstream-travelling waves. Specifically, at high Mach number, these modes become evanescent above a certain frequency, whereas at low Mach number they become progressively trapped with increasing frequency, which inhibits their reflection in the nozzle plane. |
A. Agarwal S. Phillips, P. Jordan The sound produced by a dripping tap is driven by resonant oscillations of an entrapped air bubble (Journal Article) In: Scientific Reports, vol. 8, no. 1, pp. 1-12, 2018. @article{PAJ18,This paper details an investigation into the characteristic ‘plink’ sound produced by water droplets impacting a liquid surface, such as those falling from a dripping tap. Modern high-speed video and audio capture techniques have been applied to this problem for the first time. Previous literature investigating the underwater sound produced has been validated, with the key sound producing feature both above and below the water confirmed to be the entrainment of a small underwater air bubble. Recorded sound frequencies have been shown to align with the theoretical natural oscillation frequency of the entrained bubble, confirming this to be the driver of the characteristic ‘plink’ sound. For the first time these oscillations of the entrained bubble have been directly observed on video footage. An investigation into the effect of underwater reverberation showed that the airborne sound field is not simply the underwater field propagating through the water-air interface, as had previously been assumed. An alternative hypothesis is that the oscillating bubble induces oscillations of the water surface itself, giving a more efficient mechanism by which the underwater bubble drives the airborne sound field. A model for this new hypothesis produces good agreement with experimental data. |
A. Agarwal A. L. Gregory, J. Lasenby Using geometric algebra to represent curvature in shell theory with applications to Starling resistors (Journal Article) In: Royal Society Open Science, vol. 4, no. 171212, pp. 1-13, 2017. @article{GAL17b,We present a novel application of rotors in geometric algebra to represent the change of curvature tensor that is used in shell theory as part of the constitutive law. We introduce a new decomposition of the change of curvature tensor, which has explicit terms for changes of curvature due to initial curvature combined with strain, and changes in rotation over the surface. We use this decomposition to perform a scaling analysis of the relative importance of bending and stretching in flexible tubes undergoing self-excited oscillations. These oscillations have relevance to the lung, in which it is believed that they are responsible for wheezing. The new analysis is necessitated by the fact that the working fluid is air, compared to water in most previous work. We use stereographic imaging to empirically measure the relative importance of bending and stretching energy in observed self-excited oscillations. This enables us to validate our scaling analysis. We show that bending energy is dominated by stretching energy, and the scaling analysis makes clear that this will remain true for tubes in the airways of the lung. |
Gregory, A.; Agarwal, A.; Lasenby, J. Using Geometric Algebra to understand bending in the deformation of flexible tubes (Conference) 11th International Conference on Clifford Algebras and Their Applications in Mathematical Physics, Ghent, Belgium, 2017. (BibTeX) @conference{GAL17, |
Kay, E.; Agarwal, A. DropConnected neural networks trained on time-frequency and inter-beat features for classifying heart sounds (Journal Article) In: Physiological Measurement, vol. 38, no. 8, pp. 1645-1657, 2017. @article{KA17, |
A machine-learning algorithm to grade heart murmurs and stage preclinical myxomatous mitral valve disease in dogs (Journal Article) In: Journal of Veterinary Internal Medicine, 2024. |
Unravelling the mechanics of cuff-based blood pressure measurement and the resulting sources of error (Inproceedings) In: Proceedings of the 26th International Congress of Theoretical and Applied Mechanics, pp. 581-2, IUTAM 2024. |
Neural network analysis of pharyngeal sounds can detect obstructive upper respiratory disease in brachycephalic dogs (Journal Article) In: PLOS ONE, vol. 18, iss. 8, no. e0305633, pp. 1–16, 2024. |
Acoustic resolvent analysis of turbulent jets (Journal Article) In: Theoretical and Computational Fluid Dynamics, vol. 38, pp. 687–706, 2024. |
An explanation for underestimation of systolic pressure in cuff-based measurement (Inproceedings) In: Journal of Hypertension, pp. e87, 2024. |
An empirical model of noise sources in subsonic jets (Journal Article) In: Journal of Fluid Mechanics, vol. 965, pp. A18, 2023. |
Stethoscope acoustics (Journal Article) In: Journal of Sound and Vibration, vol. 539, no. 117194, pp. 1–18, 2022. |
Detection of heart murmurs in phonocardiograms with parallel hidden semi-Markov models (Proceeding) Proceedings of Computing in Cardiology 2022, 2022. |
Aerodynamics and motor control of ultrasonic vocalizations for social communication in mice and rats (Journal Article) In: BMC Biology, vol. 20, no. 3, pp. 1–15, 2022. |
Machine intelligence for the detection of equine heart murmurs (Journal Article) In: Equine Veterinary Journal, vol. 54, pp. 15–16, 2022. |
An experimental investigation to model wheezing in lungs (Journal Article) In: Royal Society Open Science, vol. 8, no. 2, pp. 1–20, 2021. |
Narrative review of the role of artificial intelligence to improve aortic valve disease management (Journal Article) In: Journal of Thoracic Disease, vol. 13, no. 1, pp. 396-404, 2021, ISSN: 2077-6624. |
Ambiguity in mean-flow-based linear analysis (Journal Article) In: Journal of Fluid Mechanics, vol. 900, no. R5, 2020. |
A theory for stethoscope acoustics (Inproceedings) In: Proceedings of the 23rd international congress on acoustics, pp. 4991–4998, 2019, (Best Student Paper Award). |
The annoying dripping tap (Journal Article) In: Physics Today, vol. 71, no. 12, pp. 70-71, 2018. |
Jet–flap interaction tones (Journal Article) In: Journal of Fluid Mechanics, vol. 853, pp. Pages, 2018. |
The sound produced by a dripping tap is driven by resonant oscillations of an entrapped air bubble (Journal Article) In: Scientific Reports, vol. 8, no. 1, pp. 1-12, 2018. |
Using geometric algebra to represent curvature in shell theory with applications to Starling resistors (Journal Article) In: Royal Society Open Science, vol. 4, no. 171212, pp. 1-13, 2017. |
Using Geometric Algebra to understand bending in the deformation of flexible tubes (Conference) 11th International Conference on Clifford Algebras and Their Applications in Mathematical Physics, Ghent, Belgium, 2017. |
DropConnected neural networks trained on time-frequency and inter-beat features for classifying heart sounds (Journal Article) In: Physiological Measurement, vol. 38, no. 8, pp. 1645-1657, 2017. |
