MEDICAL
Improving treatment of respiratory diseases through AI requires more accurate CO2
Monitoring
Neill Ricketts, Gas Sensing Solutions (GSS) A
ccording to World Health Organisation statistics 455,000 die every year from asthma. Here in the UK, asthma kills on average three people every day, with an asthma attack happening every 10s, states NHS England. The problem is getting worse. In April, the charity Asthma + Lung UK highlighted the problem was getting worse with an increasing death toll. with professional and elite athletes actually more likely than the average population to suffer, with deaths in this group too. What makes a difference in the event of an attack is access to medicines and, if it progresses to become life threatening, an understanding of the phase of the attack, with required treatment changing as the attack progresses.
The use of AI / machine learning algorithms interest in the treatment of asthma through especially in hospitals that lack a specialist.
CO2 Measurement and Asthma Management
In the clinical setting, measuring breath CO2 concentrations allows healthcare providers to assess the stage of an asthma attack. Asthma attacks are characterised by the constriction of the air tubes (bronchi and hampers the lungs’ ability to exchange gases. By analysing CO2 levels, medical professionals can gauge how effectively a patient’s lungs are functioning and determine the severity of the attack. Traditionally, CO2 detection has been performed using capnography, a technique that provides real-time monitoring of CO2 levels in exhaled breath, displaying a 2 concentration throughout the respiratory cycle. Capnography is widely used in emergency and critical care settings to monitor patients with respiratory distress, including those experiencing asthma attacks.
The interpretation of CO2 levels, particularly during an asthma attack, is complex and requires expert knowledge. High levels of CO2
can suggest hypoventilation, a condition where the lungs are unable to expel CO2 the blood and a potential risk of respiratory failure. Conversely, low CO2 levels might indicate hyperventilation, where rapid breathing causes excessive CO2 expulsion, often a sign of a milder attack or anxiety response. Understanding these nuances is vital for timely and effective asthma management and not every hospital, especially in more remote locations, has access to these experts.
AI in Asthma Treatment: Enhancing Precision with CO2 Data Advances in AI have the power to improve how CO2 measurements are used in asthma treatment. By integrating AI algorithms with capnography data, hospitals can analyse breath CO2 levels more accurately and rapidly than ever before. AI systems can process vast amounts of data, identifying patterns that might be missed by human observation alone. This enables a more precise determination of how far an asthma attack has progressed and allows for personalised treatment strategies. For instance, AI can be trained to recognise 2 waveforms that correlate with different stages of an asthma attack. By continuously monitoring a patient’s CO2 levels, the AI can alert healthcare providers to subtle changes that indicate a worsening condition, prompting immediate intervention.
Indeed, AI algorithms have already been developed, that are able to run on smartphones and improve diagnosis, monitoring and management. Literature development of such algorithms will require both training data and future-patient data
CO2 monitoring for AI analysis Such new diagnostic requirements necessitate the waveform be captured in near real time, and at high precision, and there are several challenges to achieving a
34 NOVEMBER 2024 | ELECTRONICS FOR ENGINEERS
2
2
22
amplitude and the spatial frequency content.
Firstly, gas needs to be removed from the system before sampling and sensors need to be able to work with as small a sample size as possible, ideally sub 3 cm3
.
Next, for AI diagnostics in these fast-progressing pathways, real-time measurement is critical and a sensor sampling rate of 20 Hz or higher should be a priority. Similarly, resolution is also critical, with levels of better than 0.02 percent recommended.
Finally, while data levels will be low, the output needs to be considered, with simple universal standards such as the embedded protocol UART.
Solid-state NDIR CO2 sensors, for example the SprintIR-R, that meet these standards
and are a suitable sensor to be used are entering the market and integrating using these to train AI/machine learning improve diagnoses and treatment.
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54
