Establishing lung ultrasound as a key tool in the stratification and monitoring of COVID-19 patients
Lead Research Organisation:
University of Leeds
Department Name: Electronic and Electrical Engineering
Abstract
Lung ultrasound (LUS) is a powerful tool for the diagnosis of different pathologies of the lung. As the main cause of death in COVID-19 patients is from pneumonia, simple, low cost and effective techniques for monitoring the lungs of patients is critical. This proposal seeks to develop the necessary tools to ensure LUS can achieve this in the short and long term. The first goal of this proposal is to rectify the fact that there are currently no computer simulations of LUS for researchers to run simulations with. Implementing this model in free to use software will allow for the rapid study and optimisation of LUS by the research community. The second goal of this proposal is to implement a recently developed ultrasound beamformer for use with the range of transducers used in LUS. This novel beamforming technique has been demonstrated to improve the contrast to noise ratio and spatial resolution of ultrasound images, which enhance the detection of lung pathologies associated with COVID-19 patients. Once validated on lung mimicking phantoms and a healthy volunteer, this technique will be published in an open access journal for implementation on any other ultrasound system. The final goal is to establish a secure repository of clinical LUS images, which can be used to train deep learning networks in order to 'de-skill' the use of LUS. Furthermore, we will implement a weakly supervised deep learning network to test these datasets and those acquired from test phantoms.
Publications
Wolstenhulme S
(2021)
Lung ultrasound education: simulation and hands-on.
in The British journal of radiology
Jari R
(2022)
The diagnostic performance of lung ultrasound for detecting COVID-19 in emergency departments: A systematic review and meta-analysis.
in Journal of clinical ultrasound : JCU
Howell L
(2024)
Deep learning for real-time multi-class segmentation of artefacts in lung ultrasound
in Ultrasonics
Description | The first key finding is that COVID-19 pneumonia can not be diagnosed using ultrasound imaging alone. This is a key baseline finding to ensure that this imaging modality is implemented correctly. However, we have demonstrated that it can be used in a clinical care setting as a stratification tool. Whilst doing the training for the machine learning models we have established that there seems to be a very limited approach when labelling the data. This is particularly key when dealing with fresh datasets, such as LUS images. However, these type of techniques are becoming very prevalent in medical imaging and we feel that they should be approached with a suitable amount of caution. |
Exploitation Route | The use of ML for identifying pneumonia in the lung (not just COVID based) can be a very useful tool, and it is hoped that with the correctly trained models LUS imaging can play a key role in this. Also with the advent of 'long COVID', LUS may become a useful tool in the monitoring of these patients without the need for repeated radiation based imaging techniques. |
Sectors | Healthcare |
Description | National COVID-19 Chest Image Database (NCCID) |
Geographic Reach | National |
Policy Influence Type | Participation in a guidance/advisory committee |
Description | Clinical Partner |
Organisation | Harrogate and District NHS Foundation Trust |
Country | United Kingdom |
Sector | Public |
PI Contribution | I have coordinated both a data request for lung ultrasounds of COVID patients, and will be leading a NHS ethics application that will involve this clinical partner. |
Collaborator Contribution | Dr Martin Huntley is a Consultant in Anaesthetics and Intensive Care Medicine at Harrogate & District Hospital NHS Foundation Trust and has provided clinical input into this project. Furthermore, when we have ethics approved he will undertake bespoke patient examinations. |
Impact | Not at this stage. |
Start Year | 2020 |
Description | Commercial partner |
Organisation | Cephasonics Inc |
Country | United States |
Sector | Private |
PI Contribution | As part of this project, we are developing new beamforming algorithms for lung ultrasound imaging with patients that have COVID pneumonia. These algorithms are to be first tested on in-house ultrasound systems but will then be implemented on the Cephasonics system to help establish that they are applicable to OEM ultrasound systems and could thus be used widely. |
Collaborator Contribution | Cephasonics have provided technical support and a discounted OEM imaging system in order to test our beamformers on. They will also help implement these on systems that should go for eventual FDA approval. |
Impact | None as of yet, as the project is still in it's early stages. |
Start Year | 2020 |
Description | Presentation at the IPEM Ultrasound Update Meeting |
Form Of Engagement Activity | A formal working group, expert panel or dialogue |
Part Of Official Scheme? | No |
Geographic Reach | National |
Primary Audience | Professional Practitioners |
Results and Impact | I gave a talk at the IPEM Ultrasound Update meeting. This meeting is specifically arranged to give updates on policy and practice for use of Ultrasound imaging in the clinical setting. My aim for giving this talk is to highlight our research finding, but to start further dialogue about the use of machine learning in ultrasound imaging and how this activity might be more standardised. Specifically focusing on the acquisition and labelling of clinical images for model training. Please note that this activity has not taken place by the time this was filled out. |
Year(s) Of Engagement Activity | 2023 |
URL | https://www.ipem.ac.uk/what-s-on/ipem-events/ultrasound-update-2023/ |