TODO: Move stuff into subdirectories
TODO: Cross-link to manufacturing and other docs
The figure below displays the physical pneumatic layout of the system.
- The current pneumatic control system (Ventilator Rev 0.3) is based on a single closed-loop pressure control around the patient pressure sensor, commanded by a state machine, with blower pinch valve, exhale pinch valve, and oxygen PSOL as actuators.
- Overall structure is OxygenLoop(PressureLoop(Actuator)).
- A flow sensor observes the combined air and oxygen flow to the patient and provides an estimation of the tidal volume.
- A slow outer loop controls the oxygen mix based on the O2 sensor.
- The proposed control starting with Ventilator Rev 0.4 adds a separate flow sensor for air and oxygen input. Overall structure is Pressure Loop -> Flow Command -> Actuator. Flow becomes controlled instead of merely observed.
- Overall structure is OxygenLoop(PressureLoop(FlowLoop(Actuator))).
- Air and oxygen limbs will each have their own closed-loop flow controller, with the blower pinch valve and oxygen PSOL as actuators.
- The flow commands to these inner loops, as well as to the exhale pinch valve, come from an outer loop pressure controller which is closed around the patient pressure sensor.
- The pressure command to this outer loop comes from the state machine.
- FiO2 is controlled by a slow, outer-most control loop that changes the ratio of mixing between air and oxygen flow commands, and is closed around the oxygen sensor.
Oxygen is supplied from an external pressurized source to the oxygen port. Air is drawn in by the blower from the ambient room air. The device can also operate without an external supply of pressurized oxygen in the event of supply shortages, (delivering only 21% FiO2).
Air is drawn in by the blower through a replaceable HEPA filter and pressurized for use within the system. The blower maintains a relatively constant speed and is open-loop controlled. The capability to control the fan speed to conserve power and increase response time has not been incorporated in the current release. After leaving the fan, the pressurized (5 kPa) air passes through a check valve included to prevent oxygen backflow into the blower, as the blower is not rated for oxygen duty and could present a fire risk. Following the check valve, a proportional pinch valve regulates the downstream airflow. It does this by adjusting the cross sectional area of the flow path in the tubing built-in to the pinch-valve.
The oxygen circuit is fed from an external oxygen source, with a built-in regulator to step the pressure down to a pressure low enough to allow it to be controlled by a 12V proportional solenoid (PSOL). The design oxygen pressures at the inlet are 440 kPa - 120 kPa. The highest pressure connection that is allowable is set by the capability of the regulator, which cannot be used at a pressure higher than its rated pressure. The minimum pressure is set by the orifice size in the PSOL. With a supply pressure below the PSOL minimum, it cannot generate sufficient flow to the patient.
A check valve prevents contamination of the hospital oxygen system. Oxygen flow to the patient is controlled by the PSOL, whose orifice size is controlled by a variable pulse-width signal. Currently, the PSOL is a small automotive solenoid, which is not qualified for oxygen duty, though it is functionally suitable for development. Identifying a more reliable oxygen injection method is a high priority for improving the design. It is the understanding of RespiraWorks that the current PSOL is capable of being oxygen cleaned, and it enables fine control and rapid development (assuming) they can be sourced. As such, it has been included in the design.
The oxygen and air streams are mixed in the patient tubing without a dedicated oxygen blender; instead, the oxygen blending function is provided by proportional control of the two input streams. A galvanic cell oxygen sensor is used to measure the oxygen content as the gasses mix and provide feedback to the controller to calculate, display, and enable closed-loop control of FiO2.
After mixing, the combined gasses pass through a venturi differential-pressure flow sensor, where the flow rate of gas delivered to the patient is measured. This gas passes a medical-grade anti-viral filter upon exit from the machine for protection against viruses, bacteria, or other foreign material. If these filters are not available, the device also includes the capability to add an inline HEPA filter in the panel mount for the tubing. This feature was included due to the difficulty sourcing medical viral filters and the desire to support additional options.
The patient circuit utilized by the ventilator is an ISO 5356-1 standard 22mm female dual limb circuit which interfaces to male ports on the ventilator. If a humidifier, HME, or other device is used, these are connected to the patient inhale or exhale circuit, as appropriate.
On the expiratory path, the exhaled gas returns to the ventilator via a medical-grade anti-viral filter. This filter captures droplets exhaled by the patient. This reduces the possibility of machine contamination and dispersal of aerosolized virus from escaping into the local environment. The expiratory flow rate is measured by a flow meter similar to the one used on the supply side.
Following exhale flow measurement, another proportional pinch-valve is used to set PEEP.
Air exiting the machine is processed through a final HEPA filter to provide an additional safety factor for staff protection.
A key decision was made early on to use a blower (rather than a pressurized oxygen source and control valve), to deliver patient air.
WS7040 fan with air intake assembly
The rationale for use of a blower in general, and this blower in particular, had a few bases. First, using a blower (rather than compressed air) allows the device to provide emergency ventilation with only a relatively small power supply (around 50W continuous). This allows the device to be portable and makes the device amenable to use in field hospitals or mass casualty incidents where providing compressed air or compressors can be challenging at scale. The blower selected is also a common CPAP blower rather than a high performance ventilator unit.
This allows the use of an easily achieved blower performance specification, reduced cost and increased supply chain resilience.
However, selection of a CPAP blower does come with drawbacks—the blower is not rated for oxygen duty and the pressure response of the blower is slow, with a time constant some 1500 ms longer than comparable blowers in ventilators. This led to the development of a high-flow, long-life proportional valve, discussed in a subsequent section.
By placing oxygen mixing after the blower and using a check valve to prevent backflow, the design ensures that the blower will never see high oxygen concentrations.
Additionally, based on medical feedback, the delivered patient pressure is no more than 60 cm H20. It is common for ventilators to specify higher in order to ensure adequate ventilation for highly resistive airways . Adopting this lower pressure spec allows the ventilator to provide a pressure range that is most often needed, while avoiding significant cost. It is compliant with 80601-2-12 and all specifications reviewed for Covid-19 ventilators.
The ventilator uses a dual limb patient connector (i.e., flow is sent to the patient in one tube, and returns in a second). The flow is always one way, starting from the device going to the patient, and returning in a second limb. The other option is a single limb design, with an exhale valve at the patient side.
A dual limb design was selected to allow for mechanical PEEP control, as discussed in the next section.
Note as well that there is no check valve between the mixing volume and the patient. Backflow is prevented through two means. First, there are two check valves upstream of the tee, and no exhaust ports between the tee and the patient, so gas can only flow back into the ventilator as much as compressibility allows, which is not significant. Second, even when fully closed, the inhale proportional valve allows a small leakage (bias) flow through the device, ensuring that, as long as the device is on, flow should never enter the ventilator through the inhale port.
In a survey of suppliers, very low-cracking check valves (i.e. duckbill valves) that were available in an inline configuration were not common. The one that was tested had significant flow resonances at low velocities which interfered with volume sensing. The present approach uses no medical check valves. Normal (i.e., high-cracking) check valves can be used on the air and oxygen lines.
One of the differentiating features of the RespiraWorks pneumatic circuit is the use of a controlled mechanical valve to maintain PEEP, rather than a traditional passive PEEP valve. This decision was made early on for several reasons. First and foremost, early in the pandemic, PEEP valves were essentially unavailable and so they were not available for prototyping, but also implied they might not be available for some time. PEEP valves are effective, and so the option to design a new PEEP valve, or base a valve on other open source designs was explored. However, calibrating the spring constants proved to be a challenging manufacturing process. Generally these types of passive PEEP valves are considered disposable and so don’t experience long term use. This led to concerns about long-term use of these PEEP valves as the spring constants change with time/wear, motivating our incorporation of the controlled solution.
Incorporating a controlled mechanical valve also provides the ability to fully close the airway during inhale, thus preserving oxygen and improving the response time of the inhale circuit.
Finally, the use of a controlled valve allows the device to adjust settings based on user input, rather than requiring staff to mechanically adjust the ventilator.
A fail-open valve also prevents the need to include a separate valve for anti-asphyxia protection, though this is still being evaluated with respect to the hazard assessment..
The separate question has been raised of why not to use a pilot actuated valve, either using patient pressure or a small proportional solenoid. This was avoided entirely for sourcing considerations and was decided early in the pandemic; this could be re-evaluated for future iterations. In general, the pilot operated valves, such as those from Philips and Galemed, were deemed a risk for countries without existing supply chains to those manufacturers. The device seemed difficult to safely re-engineer.
The design process for the entire ventilator started with a desire to avoid custom components at all costs. However, after an exhaustive search of valves, control methods, fans, and spare ventilator parts, it was decided that a custom valve was the only option that would allow the design to achieve targets on cost, supply chain, and features for expiratory flow control. This decision was undertaken considering the significant quality and qualification burden required for custom parts.
A rapid pressure swing on the inhale is required to provide ventilation to low compliance lungs at a high respiratory rate, which was to that point impossible using closed-loop control of the fan alone.
The design originated by trying to tackle the PEEP problem as discussed
above. A design was specified in order to maintain a variety of
different PEEP levels and to avoid significant wasted flow during
inhale. In the process of development, it was realized that the same
design could also be used to provide a rapid pressure swing on the
inhale flow, in addition to exhale.
One of two pinch valves in each ventilator (early model)
The general operating principle of the pinch valve is to use a high-torque stepper motor to rotate a cam/rotor onto peristaltic pump tubing. The cam pinches the tubing against a platform affixed to the base. Other than the interior tubing, the valve assembly never comes into contact with the gas stream. The highest wear portion of the design is the tubing itself, which can be replaced for less than 1 USD when it must be replaced.
One option explored, not currently implemented, is to provision the device and arrange the components with an extra length of tubing (not shown above) such that it can be fed through the valve over time; pushing the extra from upstream to downstream of the valve and changing the wear point. This maintenance operation will extend valve lifetime significantly and should be able to be performed by hospital staff in addition to maintenance technicians.
Life-leader testing is currently underway to demonstrate the ultimate lifetime of the valve; though it will take time to accumulate. As of submission of this proposal, the valve has been running at a 200% duty cycle for 21 days continuously. Further information on life leader testing can be found in the 01-02 Progress Status Update.
An additional advantageous feature is that the valve is closed through a stepper motor which does not maintain its state when power is lost. The elastic modulus of the tubing, coupled with the fact that the rotor can not reach a "locked" position is sufficient to force the valve open if the power fails—the valve is normally open, and allows the exhale valve to also function as an anti-asphyxia pathway.
The pinch valve itself is built from relatively simple components, discussed in more detail in 05-01 05-01 Production Methods for Custom Components.
Proportional solenoids (PSOLs) are used in many applications (including other ventilators) for precise, rapid control of high pressure gases. The major factor arguing against the use of a PSOL in the RespiraWorks ventilator is that PSOLs with applicable characteristics are, almost by definition, produced for ventilator use, with an associated limitation on cost and availability. While some are available outside this market, those are typically not designed for use with high-pressure oxygen. High pressure oxygen requires real safety considerations, as the opportunity for fires can occur from improper lubricants, components, or errant sparks.
All that said, from the options considered and reviewed, a properly designed and sourced proportional solenoid is the cheapest, most reliable method for controlling high pressure gas flows, provided the flow rates and pressures are modest. There is still interest in improving the oxygen circuit to use a high-pressure rated pinch valve, perhaps in combination with a 2-way (rather than proportional) solenoid as part of a future cost optimization.
In the near term, a proportional solenoid was selected in order to facilitate development rather than tackle the high-pressure pinch valve design.
A secondary but important consideration is that in order to protect the patient from barotrauma, it is important for the oxygen valve to fail closed when power is lost, which is a feature of the selected proportional solenoid but not of the pinch valve.
One of the guiding principles of our design was to avoid using proprietary equipment and avoid disposables associated with each patient. This philosophy led us to develop a custom venturi-based flow sensor. It lacks some low-order accuracy, leading to reduced tidal volume accuracy, though still within spec. In return, by producing an easy-to-sterilize sensor the device can avoid proprietary and costly proximal flow sensors.
A venturi flow sensor with a differential pressure sensor used to calculate flow
Venturi flow meters are commonly used to measure the flow of gasses by measuring the change in pressure of the gas as the area is constrained. Relative to other sensor types, a venturi is the cheapest method for measuring flow rate, as it produces the largest signal to pressure drop ratio of any flow measurement solution.
RespiraWorks initially pursued numerous different flow measurement methods, including pneumatic methods such as pitot, pneumotachograph, or orificing flow sensors, which all require far more sensitive (and expensive) pressure sensors than a venturi. RespiraWorks also reviewed existing thermal mass flow sensors, which were not available during initial development. While they are costly, thermal mass sensors provide better low-noise signals at low flow rates and may in the long run prove to be a more reliable and robust solution. The issues with the sensor led to significant effort on software algorithms to improve the response. The effort that went into robust volume and flow integration will make it relatively easy to change to a different (better) sensor, but testing has shown the accuracy sufficient to control flow rate and estimate tidal volume within required accuracy.
For our ventilator, we designed venturis that can measure the high flow required, but that can also be made through widespread, low cost manufacturing processes (e.g. injection molding).
The venturi design was based around a garden fertilizer injector. Though all initial prototypes have been 3D printed, the required geometry is almost identical to a ½" garden venturi, which can be purchased at quantity for 2-4 USD / pc.
Venturis produce a large signal ratio. In ours, a maximum pressure drop of 5cm at 100 slpm produces a measured signal of 4 kPa, which can be measured with cheap, widely-available automotive differential pressure (dP) transducers.
In contrast, similar flow orifice and pneumotachograph style sensors require expensive amplified pressure sensors to measure signals in the range of just 100 Pa.
These dP measurements can then be converted into flow rate estimates based on the physics of the venturi. By placing one venturi on the inhale limb and one on the exhale limb, the flow rate going to the patient is estimated by subtracting the two flow measurements.
Based on conversations with doctors, there were two driving constraints for our oxygen sensing needs. First, the ventilator needed to be able to provide fine FiO2 control, not just a few set-points. Second, the ventilator needed to work with oxygen tanks and concentrators that were not actually 100% oxygen (e.g., improperly charged hospital oxygen tanks), since reliably pure O2 can be unavailable in many developing markets.
Based on both of these constraints, we needed an oxygen sensor to measure the actual oxygen content of the flow. We chose a galvanic oxygen sensor because they are widely available.
A downside of galvanic O2 sensors is their finite lifetimes. Therefore, we will likely add an ability to also measure oxygen flow rate, so the system is more robust to a failure of the oxygen sensor (and can sound an alarm that maintenance is required). We’re exploring whether a sensor that uses fluorescence quenching can provide the requisite sensitivity and responsiveness, as this would last for the lifetime of the device.
Filtration is critical to both keeping the ventilator from being contaminated by the patient or the environment, and keeping the patient protected if the ventilator does become contaminated.
In our system this is accomplished by dual filtering on both the inlet and exhale paths, such that for each path, there is a filter between both the ventilator and the patient and the ventilator and the environment.
HEPA filters were chosen because they are widely available and designed to filter both large and small particles. The HEPA filters are held in a custom enclosure that is designed to be injection molded. It’s easy to disassemble, clean, and reassemble, making it easy to change filters over time.
To date, RespiraWorks has not been able to identify a robust, reliable method for overpressure relief. Numerous options have been tested. All of these present failure modes, or had leakage issues which were likely more significant than their pressure protection function. One valve was identified which worked well, but the price was over $100 per valve. The current design does not possess a dedicated overpressure protection relief valve. The fan specification is such that it likely cannot result in a pressure greater than 55 cm H2O. However, the PSOL can easily generate pressures in excess of 60 cm H2O. There is some leakage at high pressure through the pinch valve, which is pressure dependent but this is likely not sufficient for patient protection. Additionally, most failure modes identified for the ventilator result in the PSOL failing closed; However, this is not sufficient from a risk consequence perspective, and this is an area requiring further address.
The device currently does not have a dedicated anti-asphyxia protection valve. Requirements on flow and flow resistance for anti-asphyxia have been included in 01-3 System Requirements. The current anti-asphyxia strategy in the event of ventilator failure is to actuate alarms (see Software) section and to fail-open the exhale control valve, which should provide an inhalation and exhalation pathway. It has not been evaluated if this pathway meets the flow resistance requirements identified, nor if this failure mode is appropriately addressed from a hazards perspective.
The RespiraWorks ventilator has developed organically over time. As the design matures, a number of upgrades are under consideration beyond the current design.
These include:
- Adding a pressure transducer prior to the patient inspiratory limb (either in addition to the one on the exhale venturi or replacing it). The benefit of placing the pressure transducer here is that if something happens to the line between the patient and the exhale portion of the ventilator (e.g. it is stepped on), the ventilator can detect such an obstruction and maintain basic safety.
- As mentioned above, adding flow sensing on the oxygen inlet circuit, to aid in FiO2 control and make the system more robust to failure of the oxygen sensor.
- Adding redundant pressure differential sensors to make it easier to detect a sensor failure.