Association of Anaesthetists/Barema Joint Research Grants

Background
Regional anaesthesia involves injecting local anaesthetic around nerves to block sensation in and movement of the body regions they supply. It can be used as a sole anaesthetic, or in combination with general anaesthesia for pain relief. It can improve outcomes after surgery, increase patient satisfaction and theatre efficiency, and shorten hospital stays. However, it is predominantly delivered by a limited number of experts, restricting patient access.

The best outcomes are achieved by using ultrasound to guide injections, which requires accurate image interpretation. However, human medical image analysis is known to be imperfect, increasing the risk of failure of the injection and needle trauma to nearby tissues.

AnatomyGuide™ is an artificial intelligence (AI) ultrasound image interpretation system, designed by Intelligent Ultrasound Limited. It produces a real-time coloured overlay for ultrasound, to highlight anatomical features and so aid the recognition of important landmarks (e.g. nerves and blood vessels). This study will use AnatomyGuide™ to assess the potential benefit of such guidance systems during ultrasound-guided injections for regional anaesthesia.

Aims
1. Assess the accuracy of an assistive AI system in identifying anatomical structures on ultrasound
2. Study the performance of anaesthetists during analysis of ultrasound images
3. Compare performance by anaesthetists during ultrasound image analysis, with and without the use of an assistive AI system

Methodology
This is a pilot study endorsed by Regional Anaesthesia UK. Data collected here will be used to understand the use of assistive AI for regional anaesthesia and, where necessary, inform additional studies to further explore the aims described above.

Ultrasound scan videos of relevant anatomical areas will be collected from volunteers. From each one, a single frame will be selected for analysis. Three experts will review the videos and still frames (including marking anatomical boundaries) to ensure suitability for inclusion in the study. The colour overlay of AnatomyGuide™ will be compared to the anatomical boundary markings of these experts.

Subsequently, anaesthetists (of varying experience/expertise) will be recruited from four centres in the UK. They will be asked to evaluate serial ultrasound videos and images, marking anatomical boundaries, with and without the use of AnatomyGuide™
Outcomes to be assessed are:
- Accuracy of recognition of anatomical structures
- Time taken
- Confidence
- Eye tracking characteristics
- Assessments of cognitive load during image analysis

Outcomes will be compared between study participants, to identify performance characteristics by level of experience. Differences associated with the use of AnatomyGuide™ will then be assessed, to gauge the potential impact of AI technology.

Expected Outcomes
1. Accuracy of the assistive AI system will exceed learners and match or exceed non-experts
2. Increased accuracy and confidence of participants, reduced time taken, more focused eye fixation patterns/reduced glance count and reduced cognitive load as expertise increases
3. The use of an assistive AI system will improve performance of participants

Implications
Assistive AI systems have potential to support ultrasound-guided regional anaesthesia performance and training, and optimise efficacy and safety of these techniques. This in turn could enhance patient access to these techniques through improving performance of non-experts.

Tracheostomies are small plastic tubes inserted into the neck. These 'artificial airways' are used most commonly in the Intensive Care Unit (ICU) for patients who are weak, or who need breathing support from a ventilator. Some Head & Neck surgical operations also need a tracheostomy. Around 20,000 new tracheostomies are inserted in the UK each year, with around two-thirds in ICU patients.

Patients tell us when they wake up with a tracheostomy, the worst thing is not being able to speak. Tubes have a balloon or cuff which 'seals off' the upper airway (nose and mouth) meaning that gas breathed in and out does not flow through the larynx (voice-box). Patients can be fully awake and yet unable to speak, lasting for days or weeks. Other communication is made worse by weakness, making it hard for patients to tell staff or loved ones how they are feeling, causing stress and worry and makes caring for patients more difficult.

Our research demonstrated that making the voice-box work by getting patents talking earlier had a positive effect on voice-box function, as well as giving patients back the gift of speech.

Our project aims to get patients talking earlier using a technique called 'Above Cuff Vocalisation' or ACV. We have developed a new prototype medical device that can safely deliver warm, moist gas for ACV. We call this prototype SEACTV which stands for Safe and Effective Above Cuff Tracheostomy Vocalisation. We are currently evaluating this device.

We want to add a new element to our testing by recording the voice that is produced with ACV and then testing out different features of the device which we think will produce a better voice. We also think that SEACTV will improve the function of the voice box and we can assess this by counting the numbers of coughs a patient produces - the more coughing, the better.

We will work with patients, doctors, nurses and speech and language therapists to record the effect of different ways of 'fine tuning' how we deliver ACV. Our assessment involves using a recording device which is attached via a 'lapel microphone' to a patient for up to 24 hours whilst they receive ACV. We can adjust the ACV to get the best voice we hear.

Researchers will analyse the voice and cough recordings and score the quality of the voice and count the coughs using computer software. This will help us work out what features give patients the best voice, and get the voice box working again.

By the end of this project, we aim to have developed a carefully tested (clinically validated) prototype medical device that is able to safely and effectively deliver ACV, helping patients to communicate and speeding up their recovery. This may allow innovative future developments such as automated feedback/control to improve voice quality, speed up laryngeal rehabilitation and recovery after critical illness.