The collection, storage, organization, and diagnostic power of multiple breathing parameters are at the cornerstone of respiratory medicine. With this project, our team will produce a DIY breath-assessing device, that will concomitantly detect lung vital capacity, exhale speed, sound, and volatile chemicals into a single system to feed an AI that will assess individual's lung health.
BackgroundWe breathe without serious interruption from birth to death. In health and normal conditions, breathing comes naturally, effortlessly and without thought. But certain respiratory diseases, like infections (COVID-19), asthma, obstructive respiratory syndrome and fibrosis, can severely change the physiology of breathing. From the sound our nostrils make when we inhale, to the suit of chemicals released in our breath, breathing parameters change when a person is ill. Collection, storage, organization and management of breathing parameters are at the cornerstone of respiratory medicine, and an essential tool for clinical trials evaluation of new medicines.
Lung Vital Capacity (lung volume - FVC) is the accepted parameter to assess respiratory health of an individual for the diagnosis and monitoring of respiratory progression. Today, FEV1 (Forced Exhale Volume at 1 second) is the single most used physiological parameter to assess a person's respiratory capacity, but this measurement is rarely deployed outside the clinics. Old fashion spirometers, the equipment used to measure FEV1, require training for their operation, and their large size prevents this device from usage outside clinics. Although new and portable spirometry solutions are on market, their high cost prevents access to a wider population, and to empower medical researchers with a versatile tool.
Along with lung volume, other parameters are used to assess lung health, like oxygenation, volatile compounds like NO in the breath, and the frequency and intensity of coughing. In spite of the great importance of all these parameters, their measurement is carried out by multiple equipments, causing a lack of comprehensive, concomitant data acquisition during a person’s single breath.
Engineering Part 1: Lung CapacityBreath parameters mentioned above are achieved through two different mechanisms and they are evaluated further using additional measurements and data analytic.
We propose the use of a Venturi's tube for measuring FVC and FEV1. Venturi tubes are methods for flow measurements that are commonly used in pipelines and settings where the flow must not be disturbed. This method involves data processing and is therefore computationally more expensive, but it is much easier to implement (Figure below)
Using SolidWorks, we tried to see how the airflow in the designed Venturi tube moves so that we can get the most ideal design according to the patient's comfort when using it and the most optimal dimensions for measuring the sensor. The purpose of airflow simulation on venturi tube was to know what range the air pressure sensor should cover, which may vary depending on the dimensions of the venturi tube. We also had to choose a sensor that could provide good accuracy to make the diagnosis more accurate while cheap. the selected sensor (MPXV5010DP) measures the differential pressure from 0 to 10 kPa. Also, its sensitivity is 4.413 mV/mm H2O which is a good sensitivity for our job. (Figure below)
To investigate the variations in ARS, we propose using two microphones in the venturi tube, to reduce noise and increase the depth of measurements by placing them in different sections of the tube. The recordings could then be pre-processed on the Arduino board before being fed into a phone for analysis using Deep and Short Long Term Memory (LSTM) networks. These networks process the audio using a visual representation referred to as Spectrogram in order to perform classification or regression. We propose a setting in which trained networks on the phone could alert the user of a change in their breathing pattern that could be caused by respiratory diseases.
Engineering Part 3: Breath Chemical CompositionLastly, we aim at measuring Nitrogen Oxide, a volatile radical (NO- or simply NO) that is released in lung alveoli space by lung-resident macrophages, as a proxy for the measurement of lung inflammation.
Part 3: Breath Chemical Composition - Short-term goal: setting sensor environment and stream data collection
NO sensors at the range necessary for breath detection (healthy individual range typically from 1 to 10 ppb) are not readily available on market. However, given the highly relevance for the biomedical space, ample literature has been produced for the development of a variety of NO detection system at the experimental level (see next section for details).
As a short-term goal into the development of an NO detection system, we decided to rely on a variety of commercially available gas sensor, to deeply the data collection system
Part 3: Breath Chemical Composition - Short-term goal: setting sensor environment and stream data collection
Though portable devices that measure NO in breath are available at market, those are expensive and do not allow for the simultaneous acquisition of lung vital capacity and NO detection. There is currently no cheaply described system to produce a NO detector. Common air-based sensor (list here) are not sufficiently selective for NO.
We undertook long investigation into NO detection literature, and compiled an openly available depository with relevant research. After pondering various options, we decided to pursue the production of a NO-specific electrochemical sensor, that has a similar design of that of an alcohol test.
Comments