A Wearable Extracorporeal CO₂ Removal System with a Closed-Loop Feedback

Abstract

Extracorporeal Carbon Dioxide Removal (ECCO₂R) systems support patients with severe respiratory failure. Concurrent ambulation and physical therapy improve patient outcomes, but these procedures are limited by the complexity and size of the extracorporeal systems and rapid changes in patient metabolism and the acid–base balance. Here, we present the first prototype of a wearable ECCO₂R system capable of adjusting to a patient’s changing metabolic needs. Exhaust gas CO₂ (EGCO₂) partial pressure is used as an analog for blood CO₂ partial pressure (pCO₂). Twin blowers modulate sweep gas through the AL to achieve a desired target EGCO₂. The integrated system was tested in vitro for 24 h with water, under varying simulated metabolic conditions and target EGCO₂ values, and in a single test with whole blood. When challenged with changing inlet water pCO₂ levels in in vitro tests, the system adjusted the sweep gas to achieve target EGCO₂ within 1 min. Control runs with a fixed sweep gas (without negative feedback) demonstrated higher EGCO₂ levels when challenged with higher water flow rates. A single in vitro test with whole ovine blood confirmed functionality in blood. This is the first step toward wearable ECCO₂R systems that automatically respond to changing metabolism. Such devices would facilitate physical therapy and grant greater autonomy to patients.

1. Introduction

Chronic lung disease causes serious, long-term disability and is the fourth leading cause of death in the United States, responsible for over 150,000 deaths each year [1,2,3]. Acute respiratory distress syndrome (ARDS) has a 30–40% mortality rate and affects 190,000 Americans each year, most recently in patients with COVID-19 [4,5]. Mechanical ventilation alleviates some of the symptoms but can exacerbate the original illness and is limited by the failing lung [6]. Artificial-Lung (AL)-based treatments, such as Extracorporeal Membrane Oxygenation (ECMO) and Extracorporeal Carbon Dioxide Removal (ECCO₂R), provide respiratory support without the drawbacks of mechanical ventilation-induced injury [7,8,9,10,11,12,13]. During acute cases, ALs allow the lung to rest and heal [9,14], and in chronic cases, ALs act as a bridge-to-transplant [10,11,15].

To increase patient mobility and thus improve the ability to exercise and rehabilitate, ambulatory ECMO systems have been developed in which the AL, blood pump, computer, battery, and oxygen cylinder are mounted to a wheeled pole [11,15,16]. By enabling ambulation and physical therapy while on ECMO, these systems significantly improve patient quality of life and survival compared to traditional non-ambulatory ECMO [17,18,19,20], while reducing the total hospital cost [21]. However, these extracorporeal systems are complex and bulky, limiting physical therapy [22,23,24]. Wearable systems under development by multiple research groups could potentially address these limitations (these systems are discussed under “Tradeoffs in Physical Size” in the discussions section) [25,26,27]. A concept diagram of a wearable ECCO₂R system under development in our laboratory is shown in Figure 1.

Nevertheless, no current systems can automatically adjust to the changing metabolic needs of the patient. Activities that increase the metabolic rate or exacerbate the disease will cause respiratory acidosis unless the system is manually adjusted. Physical therapy in particular causes rapid changes in metabolic activity. Resting or healing of the disease will likewise cause respiratory alkalosis unless the system is manually adjusted. Adjustments to these factors are limited by the availability of hospital staff, forcing a tradeoff between cost and patient health and comfort. Additionally, adjustments require periodic blood draws and analysis, causing delays between measurement of the patient status and a change in treatment. Further, it removes the patient autonomy that would otherwise be granted by a portable or wearable system. Automatic controllers have been previously investigated, but all have included components that preclude use in a long-term ambulatory application [28,29,30,31,32,33,34,35,36,37].

Previous work by our laboratory has demonstrated that CO₂ removal in an AL is a strong function of sweep gas flow [11,38]. Using this information, we developed a benchtop, laptop-controlled system that modulated the sweep gas flow rate to respond automatically and rapidly to simulated metabolic changes [39]. Subsequent work integrated all components into a battery-powered, portable package and improved overall performance metrics [25,40,41]. Here, we describe the next generation of that work: a further miniaturized, wearable, fully integrated, ruggedized ECCO₂R system. A subset of this manuscript was presented as a conference abstract at the 2023 ASAIO Conference [42].

2. Materials and Methods

2.1. Engineering Design

The design of the wearable ECCO₂R system is driven by the use cases of physical rehabilitation and ambulation.

These applications require a small and light form factor for wearable use, smaller than any other ECLS system reported (discussed in “Tradeoffs in Physical Size”). Specifically, it should be lighter than 2.0 kg and fit within a 25 × 25 × 10 cm footprint on the front of the chest. The system must be battery-powered, with a minimum 2 h of battery life to power the system for the duration of most ambulation activities. It must be ruggedized so it can continue operating after being dropped or bumped, as well as withstand minor amounts of debris. Critical components that are known to fail (battery and blowers) must be redundant.

The ability to automatically and quickly respond to changing patient blood CO₂ levels would further increase patient comfort and the ability to exercise. It must be able to maintain a stable blood pCO₂ when challenged with changing metabolic rates, particularly when compared to the current standard of care of a fixed sweep gas flow rate. To accomplish this, it should be able to maintain the partial pressure of Exhaust Gas CO₂ (EGCO₂) at a fixed value. This value should be achieved within 2 min and held steady within a range of ±5 mmHg and may be as low as 10 mmHg, which has been shown in previous studies to remove most of the CO₂ in the blood [25,40,41]. The result is that the sweep gas flow rate should be higher during periods of high metabolism and lower during low metabolism.

Since this is an ECCO₂R system as opposed to an ECMO system, it only needs to be able to process a fraction (1 L/min) of cardiac output. The system should be applicable in an arterio-venous attachment configuration or in veno-venous (VV) or veno-arterial (VA) with the addition of an external blood pump.

2.1.1. Negative Feedback Control of EGCO₂

To be able to respond to the changing respiratory needs of the patient, the system implements a negative feedback control loop, which automatically adjusts sweep gas flow rate through the AL (and thus CO₂ removal) as required by the patient (Figure 2). Ideally, patient blood CO₂ would be directly measured and used as the control signal in the negative feedback loop. However, a long-term blood-contacting CO₂ sensor does not exist. Here, EGCO₂ is used as a surrogate for blood CO₂ and controlled using a proportional-integral-controller. This is described in more detail in previous studies [25,40,41]. In this process (Figure 2), the current EGCO₂ value is read and compared to the target value (tEGCO₂). If the EGCO₂ is larger than the tEGCO₂, sweep gas flow through the AL is increased to remove more CO₂ from the blood passing through the AL, which in turn results in a lower EGCO₂. If the current EGCO₂ is less than the tEGCO₂, then the sweep gas is decreased and less CO₂ is removed from the blood.

The result is that the system can operate in a “fixed EGCO₂” mode. How well the system is able to maintain a “fixed EGCO₂” can be quantified by measuring settling time after a sudden metabolic change and the standard deviation in EGCO₂ after settling.

2.1.2. Miniaturization and Overall Design

A block diagram of the wearable ECCO₂R system is shown in Figure 3. The main components are the blower module, responsible for setting the sweep gas (room air) flow rate, the artificial lung (i.e., the “MLung”), and the exhaust module, responsible for sensing CO₂ and managing the negative feedback control loop.

To maximize comfort and mobility, a “pumpless sport-model” ECCO₂R system was chosen [38,43]. This system accesses the subclavian artery and vein to maximize ambulation and uses an arteriovenous (AV) configuration to avoid the use of a pump, both making the device smaller and more compact and avoiding pump-related lysing [44,45]. As a tradeoff, O₂ exchange is limited and it is not usable for patients that also require cardiac support.

To achieve a wearable form factor, the device must be designed around a specific AL. No other part of the system comes in contact with blood, and so, thrombosis and the immune response are almost entirely determined by the AL. The wearable ECCO₂R system is designed around the MLung, which is being concurrently developed and tested under another project [46,47]. It is an AL optimized for CO₂ removal in ESLD, making it ideal for the wearable system. It has a unique gated concentric design, allowing it to achieve similar CO₂ removal as the Novalung iLA (Xenios, Heilbronn, Germany) while using a much lower fiber surface area (0.28 vs. 1.3 m²), thereby minimizing the stagnant flow, transit time, and thrombosis. The wearable system is compatible with the multiple versions of the MLung, which have varying pressure drop, blood flow capacity, and CO₂ removal capability.

2.1.3. Blower Module

The blower module handles all safety critical processes, primarily controlling the flow of the sweep gas through the MLung. To achieve this, centrifugal WM-7040 blowers (Wei Cheng Electronics, Fenghua, China) push the ambient air sweep gas through a flow manifold and replaceable bacterial filter (Figure 4B) and into the AL. A flow sensor (Sensirion SFM4100, Stäfa, Switzerland) reads the sweep gas flow rate value. The blower module also has two ABP2 pressure sensors (Honeywell, Charlotte, NC, USA) to measure ambient pressure and the gas input pressure to control/prevent gas embolisms (i.e., gas bubbles in blood) due to high sweep gas pressures. These components are integrated and controlled by a low-power STM32F103C8T6 microprocessor (STMicroelectronics, Geneva, Switzerland).

Using feedback from the flow sensor, the blowers can be commanded to operate at a specific flow rate, allowing the blower module to act as an independent unit capable of operating in a fixed-sweep mode like existing clinical AL systems, if desired. If the blower module detects an error in the exhaust module, or if an error is indicated by the patient through a switch, it can resort to a pre-specified default fixed-sweep flow rate as in current standard of care systems. The supply of sweep gas to the AL is the highest priority (i.e., most safety critical) function of the system. For this reason, two redundant blowers and two redundant batteries were used to minimize the chance of failure of this safety critical function.

By using centrifugal blowers instead of compressed air, the system avoids the need for compressed gas tanks. This makes the system significantly more compact, and capable of being integrated into a truly wearable form factor. However, this also restricts the sweep gas to filtered room air instead of compressed O₂. Although the system can still provide some supplementary oxygenation, it is better suited to ECCO₂R applications.

With the controller on, the blower module is programmed to have a minimum flow rate of 250 mL/min, so that some air is always moving through the AL into the exhaust module. This serves two purposes. First, if flow is permitted to approach zero, the exhaust module will only measure stagnant air, rather than the air passing through the AL, and the negative feedback loop would not function correctly. Second, at low flow rates, the amount of time for the gas to flow through the AL to the exhaust module is long, increasing delay in the control loop and reducing the ability of the system to respond rapidly to changing blood CO₂ levels.

2.1.4. Exhaust (Smart) Module

The exhaust module handles all non-safety-critical processes, primarily reading and interpreting the EGCO₂ and calculating the next suggested sweep gas flow rate. To that end, it has a sensing chamber containing a Sensirion STC31 CO₂ sensor. Preceding the CO₂ sensor is a water trap, Nafion™ tubing (CO2Meter TUB-0003, Ormond Beach, FL, USA), and a heated enclosure (40 °C) to prevent condensate from reaching the CO₂ sensor, confirmed by a Sensirion SHTC3 humidity sensor. To reduce the volume of air that needs dehumidification, side-stream sampling is performed using a Thompson 2002–0215 sampling pump (USA Thompson, Sheboygan, WI, USA). A Honeywell ABP2 pressure sensor in the sensing chamber permits normalization of the CO₂ measurement. As a precautionary measure, the exhaust gas flows through a bacterial filter at the outlet of the exhaust module. These exhaust module components are integrated with and controlled by a second STM32F103C8T6 microcontroller, which also communicates with the blower module.

Physically, the overall shape of the exhaust module is 80 mm in diameter to fit under the MLung. To achieve this, the water trap is integrated into the chassis (Figure 5B) and occupies any space not occupied by the sensing system, maximizing the use of space.

In addition to these physical components, the exhaust module performs the calculations for the negative feedback loop and “suggests” a sweep flow to the blower module. The blower module can ignore this “suggestion” if it detects something wrong with the exhaust module and default to the fixed-sweep mode. Some potential faulty behaviors could include the following: an unchanging suggested flow rate indicating a non-responsive exhaust gas module, or an unusually high, low, or unchanging reported EGCO₂ value or suggested sweep gas value. As such, the exhaust module with its many sensors and negative feedback algorithm can fail without catastrophic consequences for the patient.

2.1.5. Usability Features

The two redundant battery connections were implemented to be hot-swappable, so that a near-empty battery can be swapped with a full battery while the system continues being powered by the remaining battery. This hot-swappable design allows the batteries to be charged separately from the system, minimizing the risk from battery failure while charging and exposure to wall power. Battery power is sufficient for 3 h, although larger batteries can be installed if the patient is stationary.

All physical components are mounted onto a backplate with adjustable mounting points for a harness (Figure 6). The assembled device measures 23 × 20 × 8 cm and weighs 1.7 kg with a short version of the MLung [48] and 2.0 kg with a tall version of the MLung [49]. When a patient is lying down, the harness should be removed and placed on the bed railing to avoid bedsores.

In addition to the harness carrying the ECCO₂R control system and AL, a separate “shoulder brace” (modified from [50,51]) is provided. This brace clamps to the tubing and absorbs forces that would cause decannulation or cause other strain on the cannula. Such a force may be caused by either dropping the system or the tubing being caught, with both possible scenarios in an ambulatory application. In addition, the brace keeps the blood tubing close to the patient, minimizing the risk of these events.

A basic laptop-based or tablet-based graphical user interface (GUI) and independent data logger assist with configuring the system and data acquisition during testing (Figure 7).

2.2. Subsystem Testing

The control of sweep flow and CO₂ sensing were first tested individually. The blower module was attached to the tall MLung and commanded to maintain a set of target flows. Measured flow was compared to the target flow rate.

Next, the exhaust module was directly fed known mixtures of compressed CO₂ and N₂ to ensure it could measure a range of CO₂ concentrations. Gas samples were manually taken from before the exhaust module and measured using a ABL800 Flex (Radiometer America, Brea, CA, USA). The ABL800 measurements were used as a benchmark to calibrate the exhaust module’s CO₂ readings using a one-point calibration at 20 mmHg. It was then used to evaluate the accuracy of the exhaust module’s readings after calibration.

2.3. In Vitro Benchtop Testing with Water

The wearable ECCO₂R system was tested in vitro by using it to remove CO₂ from water, an analog for how it would remove CO₂ from blood in the clinical setting. It was challenged with a range of simulated metabolic conditions via changes in incoming water dissolved CO₂ levels. The test circuit is pictured in Figure 8, showing all components of the wearable system, but deconstructed for initial testing.

The reservoir contained 5 gallons of deionized water, which is a suitable substitute for blood when only evaluating CO₂ exchange [52].

A Terumo Cardiovascular Capiox RX25 oxygenator (Terumo Cardiovascular, Ann Arbor, MI, USA) was used as a blood conditioning lung. Varying mixtures of pure CO₂ and air were pumped through the gas side of the conditioning lung to add CO₂ to the water in varying amounts to simulate changes in patient metabolic activity. Air flow to the conditioning lung was fixed at 2 L/min, and CO₂ was pumped at 0.1 L/min, 0.2 L/min, 0.4 L/min, or 0.8 L/min. Each conditioning lung setting was fixed for 4 min to allow the controller to react to each simulated metabolic change.

A centrifugal pump (3M Sarns, Maplewood, MN, USA) capable of varying water flow acted as a surrogate for the heart. A TS410 flow meter and ME-9PXL flow probe (Transonic, Ithaca, NY, USA) were used to measure water flow rates.

First, the performance of the ECCO₂R system was evaluated under different tEGCO₂s. The water flow rate was fixed at 1.0 L/min, while the tEGCO₂ varied among 10 mmHg, 20 mmHg, 30 mmHg, and 40 mmHg. For control experiments, the blower module was commanded to operate at a 1 L/min or 2 L/min fixed sweep gas flow to simulate possible sweep values that could be prescribed by a physician with a normal ECCO₂R system without active feedback. Standard deviations of various measurements were used as a metric of the system’s ability to maintain a steady-state in against a changing simulated metabolic rate.

Second, the performance of the system was evaluated under different simulated blood flow rates. The liquid pump was set to 0.5 L/min, 1.0 L/min, 1.5 L/min, or 2.0 L/min, with tEGCO₂ fixed at 20 mmHg. For these tests, the 0.8 L/min CO₂ conditioning level was omitted for expediency. As in the previous test, the EGCO₂ sensor was calibrated via ABL800 measurements prior to initiating the test. The CO₂ content in water was not evaluated because the ABL800 did not provide accurate measurements of CO₂ in water.

2.4. Blood Testing Method

In order to validate the water testing results and confirm their validity, a single test was performed using the in vitro testing setup shown in Figure 8 but using blood in place of water. The blood flow rate was set to 1.0 L/min, and tEGCO₂ was set to 20 mmHg. Air flow to the conditioning lung was fixed at 2 L/min, and CO₂ was supplied at 0.1 L/min, 0.2 L/min, or 0.4 L/min to simulate changes in the metabolic rate.

3. Results and Analysis

3.1. Subsystem Testing Results

The ECCO₂R system quickly (<2 s) attained a target sweep gas flow up to 4 L/min (Figure 9B) and exhibited a maximum sweep flow of 7 L/min with the tall MLung. Initial flow overshoot spikes were up to 0.5 L/min, but these quickly (<1 s) stabilized, at which point the error in the sweep flow averaged 0.09 L/min above the target, with a standard deviation of 0.08 L/min. The exhaust module measured EGCO₂ for all values tested up to 50 mmHg with <1 mmHg error (Figure 9A) after calibration.

3.2. In Vitro Benchtop Testing with Water Results

The ECCO₂R system demonstrated the ability to respond to simulated changes in patient metabolism to maintain a specified EGCO₂ when the negative feedback system was in effect (Figure 10). This was in contrast to the 1.0 L/min and 2.0 L/min fixed sweep gas flow (current clinical standard of care) controls, where EGCO₂ values varied widely as the simulated metabolic rate was changed. The settling time was consistently less than 1 min (

Table 1. Average, standard deviation, maximum, and minimum EGCO₂ for each test, along with settling times for fixed EGCO₂ tests. This includes periods of metabolic change.

Sweep Control	Avg EGCO2 (mmHg)	Stdev EGCO2 (mmHg)	Max EGCO2 (mmHg)	Min EGCO2 (mmHg)
10 mmHg	10.09	3.05	23.15	3.92
20 mmHg	19.43	3.15	34.38	9.24
30 mmHg	27.23	5.49	60.09	9.80
40 mmHg	30.22	9.80	73.02	9.80
1.0 L/min	16.92	17.80	57.85	2.73
2.0 L/min	4.72	5.40	20.20	0.85

). Figure 10A shows results for a 1.0 L/min water flow and a 20 mmHg tEGCO₂. Over the entire testing time, average EGCO₂ for a tEGCO₂ of 20 mmHg was 19.43 mmHg with a 3.15 mmHg standard deviation. This is in contrast to that of a 1.0 L/min fixed sweep of 16.92 mmHg with a 17.80 mmHg standard deviation and a 2.0 L/min fixed sweep of 4.72 mmHg with a 5.40 mmHg standard deviation. After settling, the average EGCO₂ with the controller on was 19.30 mmHg with a standard deviation of 1.27 mmHg. Overall, EGCO₂ with the controller on varied between 9 and 34 mmHg. For a fixed sweep flow of 1.0 L/min or 2.0 L/min, EGCO₂ varied between 3 and 58 mmHg or 1 and 20 mmHg, respectively (

Table 2. Settling time was judged as the time from when the EGCO₂ diverged from the tEGCO₂ by greater than 5 mmHg after a change in the metabolic rate, until it was within 5 mmHg for 3 s. Average, standard deviation, max, and minimum values are taken starting 120 s after a simulated metabolic change, to evaluate performance under a steady state. A more detailed chart is shown in Supplementary Information, Figure S1.

Sweep Control	Settling Time (min)	Avg EGCO2 (mmHg)	Stdev EGCO2 (mmHg)	Max EGCO2 (mmHg)	Min EGCO2 (mmHg)
10 mmHg	0.63	9.92	0.70	12.61	6.99
20 mmHg	0.51	19.30	1.27	22.16	14.72
30 mmHg	0.58	29.97	0.97	32.42	26.52
40 mmHg	0.53	39.92	0.90	42.53	37.48

). Additional metadata for all tests are supplied in Supplementary Information Figure S1.

Figure 10B displays measured sweep gas flows for the experiment shown in Figure 10A. In Figure 10B for the fixed EGCO₂ test (feedback on), sweep flow increased with increasing simulated metabolic rate and vice versa. When in fixed sweep mode, the ECCO₂R system maintained the target sweep flow with gas flow standard deviations of 0.048 L/min at 1.0 L/min and 0.082 L/min at 2.0 L/min.

3.2.1. Varying tEGCO₂ Results

For the next test, the control system was set to a tEGCO₂ of either 10, 20, 30, or 40 mmHg, and incoming water CO₂ content was varied to test the control system (Figure 11). For a tEGCO₂ of 10 mmHg, the average EGCO₂ was 9.92 mmHg with a 0.70 mmHg standard deviation once settled. For a tEGCO₂ of 30 mmHg, at the lowest simulated metabolic rate, the measured EGCO₂ was less than the tEGCO₂ at the system’s minimum sweep gas flow rate, which averaged 0.35 L/min, and thus, the tEGCO₂ could not be achieved. For the other simulated metabolic rates, the average EGCO₂ was 29.97 mmHg with a 0.97 mmHg standard deviation. For a tEGCO₂ of 40 mmHg, tEGCO₂ could not be achieved for the smallest two simulated metabolic rates; the minimum sweep averaged 0.32 L/min. For the other metabolic rates, the average EGCO₂ was 39.92 mmHg with a 0.90 mmHg standard deviation.

Average overshoot and undershoot were lowest at 7.99 mmHg for a tEGCO₂ of 10 mmHg with a 3.24 mmHg standard deviation, increasing to 10.61 mmHg with a 2.91 mmHg standard deviation at 20 mmHg, 17.88 mmHg with a 8.45 mmHg standard deviation at 30 mmHg, and finally 25.85 mmHg with a 10.14 mmHg standard deviation at 40 mmHg.

3.2.2. Varying Water Flow Rate Results

The response of the ECCO₂R sweep gas control system to changing water flow rates is displayed in Figure 12, demonstrating the ability to maintain a tEGCO₂ of 20 mmHg across a range of flows. For the range of the water flow test, average EGCO₂ with the controller on (fixed EGCO₂) ranged from 20.19 mmHg to 20.41 mmHg, with standard deviations from 3.36 mmHg to 4.50 mmHg over the course of the run. In contrast, 1.0 L/min fixed sweep gas flow controls (controller off) had average EGCO₂ values ranging from 13.24 mmHg to 43.39 mmHg, with standard deviations ranging from 12.61 mmHg to 24.30 mmHg. The 2.0 L/min fixed sweep gas flow controls had average EGCO₂ values ranging from 3.43 mmHg to 15.09 mmHg, with standard deviations ranging from 2.02 mmHg to 12.34 mmHg. After settling, for water flows of 0.5 L/min, 1.0 L/min, 1.5 L/min, and 2.0 L/min and a fixed EGCO₂ of 20 mmHg, average EGCO₂ was 20.05 mmHg with a 1.59 mmHg standard deviation, 19.79 mmHg with a 1.09 mmHg standard deviation, 20.02 mmHg with a 0.75 mmHg standard deviation, and 20.03 mmHg with a 0.90 mmHg standard deviation, respectively.

Overshoot and undershoot of EGCO₂ when targeting fixed EGCO₂ varied between 10.92 mmHg and 12.61 mmHg with standard deviation varying from 1.32 mmHg to 2.63 mmHg. Neither the average or standard deviation of overshoot and undershoot appeared to correlate with changing water flow rates.

3.3. In Vitro Benchtop Testing with Blood Results

To validate the in vitro water experiments, a single test with whole blood was performed. tEGCO₂ was set to 20 mmHg, and incoming blood CO₂ was varied (by varying the concentration of CO₂ in the sweep gas of the conditioning lung) to simulate changes in patient metabolic activity. The control system maintained the desired tEGCO₂ under all simulated metabolic rates (Figure 13), demonstrating an EGCO₂ maximum overshoot of 2.6 mmHg, maximum undershoot of 2.7 mmHg, and a standard deviation of 0.7 mmHg throughout the course of the study, including periods of changing metabolic rates.

4. Discussion

4.1. Analysis of Performance

Blower accuracy is 0.008 L/min when amortized over 10 s, although raw readings from the mass flow sensor were quite noisy at 0.25 L/min. The short MLung has half the gas-side resistance of the tall MLung and so is anticipated to have a maximum sweep of approximately 14 L/min.

Figure 12 demonstrates that the system is unable to add CO₂ to blood and only remove it.

4.2. Mechanism of Action

Gas flow increased with an increasing simulated metabolic rate when operating in fixed EGCO₂ mode. This allows the controller to maintain a fixed EGCO₂ by increasing sweep gas flow during periods of high simulated metabolic rates to remove additional CO₂ from the blood and vice versa.

It should be noted that as the gas flow rate increases, the time for pCO₂ levels in the gas to equilibrate with pCO₂ levels in the liquid decreases, and therefore, the difference between these two values is expected to increase as the sweep gas flow increases. Therefore, the effect of an increasing gas flow rate is expected to decrease the EGCO₂ down to its targeted value but pCO₂ in the blood or water to a lesser extent.

4.3. EGCO₂ Overshoot during Simulated Metabolic Changes

The control system exhibited overshoot after each simulated metabolic increase in the in vitro water tests and undershoot after each simulated metabolic decrease. The over/undershoot varied between 7.99 mmHg at a tEGCO₂ of 10 mmHg to 25.85 mmHg at 40 mmHg, which consistently settled to within 5 mmHg of the target value in less than 1 min. These overshoots were mitigated slightly by adjusting the negative feedback algorithm to be less aggressive. (Specifically, the proportional term of the proportional-integral-derivative controller was reduced, reducing overshoot at the cost of initial response time).

In the human body, blood completes a full cycle through the body in 10 to 120 s [53], approximately similar to the settling time of the control system. Further, changes in blood CO₂ in humans are typically gradual and caused by rest/wake cycles, disease progression or healing, and changes in patient activity. The sudden step changes in incoming water and blood CO₂ levels used in the in vitro tests were not fully realistic, but instead were employed to provide a difficult challenge for the control system and to ensure the control system did not resonate. Finally, any over/undershoot of tEGCO₂ with the controller on was still significantly better than variation in the EGCO₂ with the controller off (fixed sweep flow).

4.4. Inability to Perform Low CO₂ Evacuation

While testing various tEGCO₂ values (Figure 12), tEGCO₂ settings of 30 and 40 mmHg could not meet the tEGCO₂ value for the smallest simulated metabolic rates. In these instances, the measured EGCO₂ was below the target value at the smallest permitted sweep gas flow rate. Since the sweep gas flow could not go lower, EGCO₂ could not increase any more. In patients this should not be an issue, given that the system is meant for patients who have poor CO₂ removal and thus high blood CO₂ levels.

In fact, an extended period of fixed sweep flow in patients can sometimes lead to low blood CO₂ levels (hypocarbia) and its associated complications (cerebral vasoconstriction, cardiac arrhythmia, and exacerbation of existing cardiorespiratory conditions). For a patient on a ventilator, hypocarbia can only be determined via periodic blood draws, and thus, detection (and an adjustment of sweep flow) may be delayed. The system described here should help prevent hypocarbia by automatically reducing the sweep gas flow when blood CO₂ (and thus EGCO₂) decreases.

4.5. Regulation of EGCO₂ Rather than Blood CO₂

In the presented wearable ECCO₂R system, measured EGCO₂ is used as the control signal for the control system. In this case, EGCO₂ serves as a surrogate for directly measuring arterial blood pCO₂ content (PaCO₂), since no long-term blood CO₂ sensor exists. However, being able to measure and regulate blood pCO₂ directly would have been the superior choice, if possible.

EGCO₂ is a function of both the blood pCO₂ entering the AL and the sweep gas flow rate. Theoretically, a sufficiently low sweep gas flow rate would equilibrate gaseous CO₂ levels with the blood pCO₂, such that the EGCO₂ and the post-AL blood pCO₂ are effectively identical. As the sweep increases, the sweep gas has less time to equilibrate, and so there is a greater gap between post-AL blood pCO₂ and EGCO₂. In the context of this study, even though the controller maintains a constant EGCO₂ value, the post-AL blood pCO₂ likely exceeds the EGCO₂. Further, higher metabolic rates result in higher sweep gas flow rates and therefore a higher post-AL blood pCO₂. This relationship has been demonstrated in vitro [25] and in vivo [41] in previous studies that also used a fixed EGCO₂ method.

Additionally, since this device is designed for subclavian AV ECCO₂R, it processes a fraction (~10–20%) of the circulating blood. A sicker patient would have a higher pCO₂ in the blood not processed by this device, and so the blood leaving this device would need to have lower pCO₂ to compensate during remixing. Using the fixed EGCO₂ method, such a patient would be prescribed a lower tEGCO₂ than a healthier patient.

With that said, a fixed EGCO₂ sweep gas control system is still superior to fixed sweep gas. Firstly, if a physician was present, they would prescribe increased sweep during periods of increased metabolism and less sweep during periods of decreased metabolism, which is exactly how the presented controller responds in Figure 11. Secondly, higher metabolic levels should result in a higher PaCO₂ than lower metabolic levels, even with fixed EGCO₂. Therefore, the system is undercompensating rather than overcompensating. In conclusion, the performance of the negative feedback system likely falls between a fixed sweep system and a theoretical system, which would monitor and directly control blood CO₂ (such as a physician constantly taking blood-gas readings).

In practice, the rate at which a physician can change sweep is limited not only by their availability but also by the rate of blood draws, which a carry risk of infection and anemia.

We envision that a physician would prescribe a tEGCO₂ based on the patient’s baseline PaCO₂, with a lower tEGCO₂ corresponding to sicker patients. The negative feedback system would then reduce PaCO₂ variations during changes in activity and the disease state that take place in between blood draws and physician inspections.

4.6. Tradeoffs in Physical Size

This system is smaller by an order of magnitude than any other clinical ECMO or ECCO₂R system to date. Other “wearable” ECMO systems only have the AL and blood pump (if it exists) in a wearable form factor but rely on a separate package to house the gas supply and control system [26,27,54,55,56].

When considering weight, the use of blowers using ambient air rather than compressed O₂ had the most significant impact. Portable O₂ concentrators weigh upwards of 800 g, and systems with higher flow rates weigh closer to 2 kg [57]. Portable O₂ tanks for short-term ambulation weigh 1.36 kg and would last 2.75 h at 1 L/min [58]. These would have roughly doubled the weight of the system. In comparison, the two blowers used on the reported system weigh only 90 g each.

When considering space, the greatest impact was in the geometric packaging of all the components. In particular, the control electronics were built around the other components to maximize use of space and minimize overall size. However, this restricts the choice of AL to devices with similar dimensions to the MLung (without physical modifications to the system).

The sensor module weighs approximately 100 g and adds 80 mm to the bottom of the MLung. Although not negligible, its presence does not significantly impede wearability, while the advantage of negative feedback on CO₂ removal addresses changing metabolism during physical rehabilitation.

A summary of advantages and disadvantages and comparisons with other ambulatory ECMO and ECCO₂R systems is shown in

Table 3. Comparison of the current system with similar ECMO and ECCO₂R systems.

Feature	Wearable Smart ECCO2R	Portable Smart ECCO2R [25]	Breathe [26]	Wearable Pumping AL [27]
High FiO2	No	No	Yes	Yes
Compatible with multiple ALs	No	Yes	No	No
Wearable	Yes	No	With separate gas supply and controller	With separate gas supply and controller
Active negative feedback	Yes	Yes	No	No
Blood pump (for VA or VV operation)	No	Yes	Yes	Yes

.

4.7. Planned Future Studies

Future studies are planned in which these in vitro water studies are replicated in vitro with blood. During these blood studies, pre-AL and post-AL blood pCO₂ values will be measured. It will then be tested in vivo in a sheep model.

5. Conclusions

A wearable ECCO₂R system capable of automatically responding to changes in blood pCO₂ levels has been described. Specifically, it is able to operate in a “fixed EGCO₂” mode. Additionally, it is smaller and more portable than any other complete ECLS or ECCO₂R system reported. It serves as the blueprint for a system that would allow ECCO₂R patients to be treated with fewer resources, and better comfort, autonomy, and outcomes. In particular, it could reduce many of the logistical challenges associated with physical therapy while on ECCO₂R.