PhD studentship. Come and work with me! Advert is now out. See here
The studentship is only open to UK/EU students, because the funding available is insufficient to cover international fees. Applicants will need a 2:1 degree or higher, and to have skills in a numerate discipline like engineering, ideally with signal processing skills and experience in use of Matlab.
The working title is Automated Analysis of Lung Sounds as a Predictor of Ventilator Associated Pneumonia. This work is being funded jointly by the NIHR Southampton Respiratory Biomedical Research Unit and an asthma charity called AAIR. You will be supervised by Prof Anna Barney (Faculty of Engineering and the Environment (FEE)), Prof Mike Grocott (Faculty of Medicine), and me – Prof Anne Bruton (Faculty of Health Sciences). We are all three based at the University of Southampton and you will be registered with FEE for the duration of your PhD.
Some background for this PhD
I have blogged about lung sounds before. See here, here and here. Standard lung auscultation provides an assessment of airway patency and geometry, and parenchymal normality, but the information it gives is subjective and qualitative. Computer aided lung sound analysis (CALSA) removes the subjectivity and allows quantification of sounds and their characteristics (intensity, frequency) (Earis & Cheetham 2000). We hypothesise that information derived from CALSA will aid the early detection, diagnosis and monitoring of acquired pulmonary complications. Most research in this field has been conducted in controlled, laboratory environments using multiple sensors, but we have established that reliable lung sound data can be recorded with a single sensor in an outpatient clinical environment (Marques et al. 2009), and we have current PhD students who have successfully collected lung sound data from inpatient ward environments.
Acquired pulmonary complications (atelectasis, infection, pneumonia) are common in several patient groups, particularly in post-operative and intensive care populations. Mechanically ventilated patients are at high risk for developing nosocomial pneumonia or tracheobronchitis. In general, the frequency of infection increases with the duration of mechanical ventilation, but the risk appears to be greatest in the first week of intubation (Ahmed 2001). Ventilator acquired tracheobronchitis (VAT) represents an intermediate process between lower respiratory tract colonisation and ventilator-associated pneumonia (VAP), which is an important cause of morbidity and mortality with a significant economic impact on health care resources, despite advances in antimicrobial therapy and better supportive care (Rotstein et al. 2008). The incidence of VAP has been reported to range from 10% to 20% of intensive care unit patients (Rewa 2011). It is suggested that there is a short therapeutic window between the development of VAT and the progression to VAP, but one of the challenges to optimal management is having an early and accurate diagnosis, as delays in therapy are associated with increased mortality. Formal diagnosis of VAP usually comprises clinical signs and symptoms, microbiological data, and radiographic findings; but there is currently no consensus on diagnostic criteria for VAT (Craven et al. 2009).
Lung sound data are simple to collect (noninvasively at the bedside), using a digital stethoscope, and can provide data for quantitative and objective tracking of changes in respiratory health status. In this PhD, data from a cohort of ventilated patients in an intensive care setting would be collected. Data would be acquired at multiple time-points, from admission to discharge (or death) together with a clinical and pathological characterisation of the progression of any acquired pulmonary complications. Chest x-rays are routinely, and frequently, taken in the ICU environment for clinical need (e.g. checking placement of endo-tracheal and naso-gastric tubes, checking central line placement, monitoring clinical changes), so there will be opportunities to compare acoustic with radiological findings.
We plan to start initial clinical data collection in advance of the enrolment of the PhD student. The PhD student would be based in the Faculty of Engineering and the Environment and focus on acoustic analysis of lung sound recordings to establish and evaluate the usefulness of extracted acoustics features (energy, energy by frequency band, spectral tilt, temporal and spectral features of adventitious sounds such as wheezes and crackles, statistical features, cepstral analysis, MFCC and PLP-CC etc). These are all analysis methods that have been used effectively to characterise other kinds of audio, in particular speech recordings. Their use in pulmonary analysis is at an early stage due to slow embedding of the use of digital stethoscopes in clinical practice and the difficulty of establishing ground truth data for in vivo recordings. They have, however, shown promise as methods for analysing recordings from patients with cystic fibrosis (Marques et al 2012), bronchiectasis (Marques et al, 2009), COPD (Bennett et al, 2014), asthma and several other lung pathologies. Feature selection techniques will be used to identify the subset of acoustic features most sensitively predictive of VAP onset and progression for use in patient monitoring. These will then be incorporated into an automated system for identifying patients at risk of developing VAP in order to provide diagnostic support for clinical staff.
The PhD student will spend the first 9 months familiarising him/herself with the literature on lung sounds and their analysis, and establishing a set of acoustic features that can be extracted from the data sample and which have the potential to change with disease onset and progression. The next phase of the work from 9 to 18 months will be testing the statistical sensitivity of the features to known changes in patient respiratory health status to establish which should be selected as part of a predictive system. The final phase of the work will be a prospective study to test whether onset of VAP can be reliably predicted by the selected features, which will occur in parallel with thesis preparation.
References
Ahmed QA, Niederman MS (2001) Respiratory infection in the chronically critically ill patient: ventilator-associated pneumonia and tracheobronchitis. 22(1) 72-85
Beck R et al. (2007) Computerized acoustic assessment of treatment efficacy of nebulized epinephrine and albuterol in RSV bronchiolitis. BMC Pediatr. 7:22
Bennett et al. (2015) The relationship between crackle characteristics and airway morphology in COPD. Respiratory Care. 60 (3): 412-421.
Craven DE et al. (2009) Ventilator-associated tracheobronchitis: the impact of targeted antibiotic therapy on patient outcomes.Chest135:521-8
Earis JE, Cheetham BMG (2000) Current methods used for computerized respiratory sound analysis. Eur Respir Rev 10:586-590.
Marques A et al. (2009) The reliability of lung crackle characteristics in cystic fibrosis and bronchiectasis patients in a clinical setting. Physiological Measurement. 30:903-912
Marques et al. (2012) Are crackles and appropriate outcome measure for airway clearance therapy? Respiratory Care. 57(9), 1468-1475.
Murphy R.L. (2008) In defence of the stethoscope. Respir Care. 53(3):355-69
Rewa O, Muscedere J. (2011) Ventilator-associated pneumonia: update on etiology, prevention, and management. Curr Infect Dis Rep. 13(3):287-95
Rotstein C et al. (2008) Clinical practice guidelines for hospital-acquired pneumonia and ventilator-associated pneumonia in adults. Can J Infect Dis Med Microbiol. 19:19-53.
Anne Bruton's Blog 