BAMBI (a Biointelligent Apnea Monitor for Bradycardia-prone

BAMBI addresses apnea of prematurity through our tripartite system: a machine-learning algorithm to detect apneic events, a mechanical device that automates apnea resolution, and an event recording system for physicians.

About half of all premature infants in neonatal intensive care units (NICUs) experience apnea of prematurity, or periods of stopped breathing greater than 15 seconds, resulting from an underdeveloped nervous system. Among more premature infants, the incidence of apnea only increases. If apnea is unaddressed, slowing of the heart (bradycardia) may occur, leading to decreased oxygen saturation (hypoxia) and brain blood flow, and eventual infant death. The current apnea management workflow is subjective, un-optimized, and time-consuming for NICU nurses, residents, and attending neonatologists. While caring for multiple infants simultaneously, nurses must decide if an apneic event is significant and whether stimulation is required based on predetermined thresholds of blood O2, heart rate, and respiratory rate. Meanwhile, the nursing profession continues to face increasing shortages due to an aging population, nurse burnout, and lack of potential educators. These shortages have been made more apparent due to the COVID-19 pandemic. NICU nurses typically care for 3-6 infants during their shift, rather than the recommended 1-2. Nurses must also provide information to attending physicians about apneic events so they can determine an infant’s discharge timeline. EMRS (electronic medical record systems) typically relay a range of daily biological signals, which oversimplifies the data and does not provide an accurate account of each infant’s condition. Overall, the burden of work placed on NICU nurses is very high. The treatment itself for apnea of prematurity involves subjective levels of physical stimulation from the nurse until the infant is awake. When an infant stops breathing, nurses typically wait 15-20 seconds to allow the infant to self-recover. If the infant fails to recover, nurses apply gentle physical stimulation. If unsuccessful, nurses move on to more extreme methods of assisted ventilation (CPAP, NIPPV) which often cause intubation injuries or other unintended harms to the infant. Caffeine, a central nervous system stimulant, has also been used to successfully reduce apnea’s incidence. However, drug activity post-administration is relatively uncontrolled, and extended caffeine exposure can have adverse effects. Existing technologies for automating stimulation use simple thresholding methods which fall short due to their lack of adaptability, or provide continuous stimulation, which can cause infant desensitization to the stimuli over time. We aimed to detect apnea more precisely and only provide stimulation as needed, following common NICU protocols. We also uniquely aimed to integrate our device with existing NICU technologies and reporting mechanisms. BAMBI, our Biointelligent Apnea Monitor for Bradycardia-prone Infants, automates and streamlines the information reporting, infant stimulation, and significant event determination processes. BAMBI takes in patient biological signals, determines the presence of apnea through a machine learning (ML) algorithm, and uses this information to control a stimulation device. By automatically stimulating infants upon onset of apnea, collecting and transmitting data to physicians, and alerting NICU nurses if the apneic event does not end, we aim to reduce the frequency of manual infant stimulation and increase the available data. By integrating this system with existing infant monitoring systems, we aspire to increase the standard of care for premature infants experiencing apnea.

Our project has three primary objectives: 1) reduce the subjectivity in apneic event identification, 2) automate the first and second lines of infant stimulation, and 3) facilitate better data reporting to physicians. As a secondary objective, we also wanted our system to integrate with commonly seen NICU setups. As stated above, our system, BAMBI, accepts vital signs and other biological signals from premature infants and feeds that data into a pre-trained ML algorithm, which then determines if an apneic event is occurring (Figure 1). Currently, the vital signs are gathered from small sensors which send information to BAMBI via a microcontroller. The machine learning algorithm wirelessly activates infant stimulation via the mat when apnea is detected and adds the event to a report of apneic over time for physicians to use as they see fit. This workflow helps standardize the determination of what is an apneic event and relies less on subjective, in-the-moment decisions by nurses. When an apneic event occurs, we stimulate an infant in two steps: first, using vibrational motors to provide cutaneous stimulation, and second, using a linear oscillating motion to rock the mattress back and forth to mimic a stronger proprioceptive stimulation. We also alert NICU nurses if the stimulation does not terminate the apneic event, requiring further intervention. The automation of stimulation reduces the amount of time NICU nurses need to spend on resolving apneic events. Finally, we will have a user interface that will allow for ease of data transmission to attending neonatologists and neonatology residents, increasing the amount of data available for them to make discharge timeline decisions. Based on stakeholder interviews, BAMBI will be able to integrate well with current NICU layouts; the device fits unobtrusively into a premature isolette and uses wireless WIFI activation from a hospital computer.

In the short-term, our design will reduce NICU nurses’ workload and therefore help combat issues related to nurse burnout that have become more prevalent especially after the COVID-19 pandemic. BAMBI will also help improve patient care in NICUs that are chronically understaffed, and make decision-making and data-reporting easier for nurses and physicians. In the long-term, our device could have global, economic, and societal impacts (we do not anticipate many major environmental impacts as we are not reducing any existing systems’ environmental footprint). Globally, BAMBI could be used in low-resource settings where nurses are not available to monitor premature infants 24/7. BAMBI’s low cost (see Appendix B) makes our device accessible to hospitals in these regions. Economically, our device will save nurses time, meaning they can take more time in completing their other daily tasks. Ideally, this will lessen nurse burnout and decrease the financial strain that nurse shortages cause on the medical system. Societally, a decrease in burnout will also help the nursing profession and encourage more nurses to explore working in NICUs. BAMBI may also lead to a push for more standardization in preterm infant care; while guidelines for care do exist, they are often vague and do not take advantage of ML for standardization.

Our team consists of members with a diverse skillset, which have helped us greatly in furthering this project. As bioengineering students, we’ve gained experience in product development, prototyping and fabrication with laser cutting and 3D design, signal analysis, circuitry, and a basic understanding of the biomedical field, all through our classes and project-based labs. Several of our members also have background in computer programming and have previously worked with machine learning through classes and internships, which gave us direction in the software portions of our project. Another one of our members, who is on track to become a medical physician, gained important insight into the state of medicine through shadowing surgeons in her praeceptorial course, which helped us greatly with the needfinding step of our project. With our skills combined, we were able to develop our innovative solution, BAMBI. Our team members are all members of the Penn community, which encompasses not only the students at the undergraduate and graduate schools, but also the various hospitals nearby, such as the Hospital at the University of Pennsylvania and the Children’s Hospital of Philadelphia. After talking to various nurses and neonatologists at these facilities, we’ve gained a great appreciation for their work in keeping this community safe. Thus, we wanted to create something that can give back to these integral members of our community.

In developing BAMBI, we have and continue to collaborate closely with a nurse practitioner who works in two major Philadelphia NICUs (Hospital of the University of Pennsylvania, and Children’s Hospital of Pennsylvania), with whom we have regular meetings to consult about our project’s design and progress to date. We have also conducted many stakeholder interviews, including with other nurse practitioners, NICU nurses, and neonatologists about how our device can be improved to benefit each of their roles in the NICU setting. In performing these interviews, we made sure to interview nurses at different hospitals to investigate the state and setup of the NICU in different hospital environments and see how our device can best fit in the apnea treatment process. For safety and efficacy validation, we plan to survey nurses and attending physicians to gauge the device’s performance in a NICU setting prior to gaining IRB approval for physical device testing.

As mentioned above, the current predominant treatment for apnea of prematurity involves multiple levels of subjective physical stimulation by nurses until the infant wakes. Each nurse has their own methods of stimulation, and stimulation intensity and frequency can vary from NICU to NICU, resulting in a very non-standardized treatment across hospitals and nurse shifts. Continuous caffeine dosages have been demonstrated to be effective at reducing the incidence of apnea in premature infants, but it also is accompanied by adverse side effects, especially upon discontinuation. Finally, other automated stimuli either are not adapted to each infant individually, or provide continuous stimulation which can lead to infant desensitization and lower effectiveness over time. BAMBI can detect apnea more precisely than simple thresholding methods used by current automated stimulation devices due to our employment of machine learning, which also means that BAMBI can adapt to each infant and deliver better results over time. BAMBI also records apneic events in significant detail, which physicians can access; this is a unique feature that we have not seen in an existing product or process. Finally, our workflow is standardized across each infant, making our system easy to learn and operate. As such, BAMBI is a unique all-in-one system, providing infant-adaptable detection, automated stimulation, and extended apneic event data reporting, filling a niche that no currently existing solution does.

One year: Bring a reliable and integratable apnea detection algorithm and stimulation device for NICUs to clinical testing and partner with 5 nearby NICU’s for further project development. Additional steps needed to achieve these goals Work towards implementing the mechanical component of BAMBI in a clinical trial setting through an IRB Protocol Work through NICU bureaucracy to optimize how we can integrate BAMBI in the NICU environment Introduce our GUI to NICU computers for expanded physician data access and apnea tracking

Our central program is housed in Python and runs the machine learning algorithm (both training and testing), as well as the mechanical components which are currently actuated via a microcontroller, and a GUI that displays the device’s current status and apneic episode data logs (which nurses and physicians are able to interact with). Our machine learning algorithm uses random forest as our decision-making algorithm, with sensors sending information about heart rate, blood oxygen saturation, respiratory rate, and temperature to the algorithm. The machine learning component has been trained using labeled and de-identified patient data from the University of Edinburgh, and we are in constant contact with researchers across the country seeking additional data to use for our model’s enhancement. The linear actuation for the rocking stimulation and the vibrational motors for cutaneous stimulation are also controlled through the central Python program and are activated when the machine learning component detects that an apneic episode is occurring.

While our solution is still under development, we expect that our apnea detection algorithm and our data interfaces can reach NICUs in our local area within the year as part of our clinical trial process. Ideally, we will partner with the nursing team in the level IV nursery at the Children’s Hospital of Philadelphia (CHOP), allowing our devices to assist their team of 20 surgical nurse practitioners. The NICU at CHOP is quite large, with about 100 beds and an average daily census of 100 infants. We expect that a majority of preterm infants will experience significant periods of AoP based on our conversations with nurses in CHOP’s NICU. Within the next year, we hope to make 5-10 devices available to the nurses to assess their impact on workflow and any necessary future adjustments.

Highlight specific financial, technical, legal, cultural, or market barriers that may limit your impact in the next year. We currently need IRB and FDA approval in order to explore the use and testing of our system in NICUs, especially the mechanical component. Additionally, in order to fully integrate our device within the NICU setting, we will need access to proprietary software from major vital sign monitoring companies (GE, Philips, etc.) in order to be able to read data from these monitors into our algorithm.

We will provide our device to customers (being hospital NICUs), who will then use our device to improve patient care and satisfaction levels. By reducing the amount of time infants spend in NICUs (through more consistent tracking of apneic events), hospitals will be able to serve more patients, increasing their revenue. Additionally, BAMBI’s ability to lower nurse burnout will decrease nurse turnover and the increased expenses that result from high turnover and low nurse availability that burnout causes. Physicians will also save time using BAMBI because of its improved data access features, allowing them to see more patients, which again increases hospital revenue.

In the short term, we expect to use an innovation grant, such as an NIH grant, to cover the final development and testing costs of BAMBI before we can bring our product to market. Once we have ensured that BAMBI meets all the necessary federal and local standards and regulations, we will raise investment capital to scale up our manufacturing process for distribution and hospital-system-specific device customizations, after which we will sell BAMBI as a software + hardware package to hospitals for use in their NICU. Any additional revenue will be put towards improving BAMBI and developing new technologies that can be applied towards enhancing NICU efficiency, which we can then add to the BAMBI software + hardware package to sell to hospitals.