Advancing next-generation pulse oximeters and AI-assisted health tools

Related program: Diagnostics

Improving access to multimodal, AI-integrated, and smartphone-based health tools that enable providers to deliver more integrated screening, diagnoses, and care when and where it is needed

Achieving the ambitious maternal and child health targets set for 2030 demands urgent attention to equitable access to innovative healthcare solutions in resource-constrained settings. Where healthcare providers are already overburdened with responsibilities and patient loads, technology can and should do more to support providers in doing their best work. Aligned with current trends in high resource settings, health technology that integrates and automates functionalities allows providers to prioritize their time and expertise on patient interactions.

Multimodal (or next-generation) pulse oximeters—noninvasive handheld devices that expand the features of standard handheld pulse oximeters by additionally measuring respiratory rate, temperature, and/or hemoglobin—are a promising technology that can provide objective measurements to support clinical decision-making. Integrating multiple clinical measurements into a single device improves accuracy, efficiency, and patient adherence during consultations, facilitating more precise diagnoses and comprehensive illness management. Beyond reducing mortality rates, multimodal pulse oximeters have the potential to optimize resource allocation and alleviate strain on healthcare systems. They achieve this by providing critical health information, strengthening referral decisions, and reducing unnecessary hospitalizations, intensive therapy, and excessive antibiotic treatments.

Smartphone sensor-enabled health measurement as part of medical device technologies is rapidly progressing from theory to practice utilizing comprehensive health datasets to train and validate artificial intelligence (AI) models for digital health applications leveraging off-the-shelf features of readily available smartwatch and smartphone platforms. There is a critical gap in the reach of medical devices in low-resource settings where supply of basic health technologies like pulse oximeters, hemoglobinometers, and patient monitors do not meet clinical demand, while many companies prioritize wellness products for high-income populations. With a market penetration of more than 6.8 billion users, smartphones could be the most readily available health tool already in the pockets of most providers, health workers, and patients globally. To achieve this, there is a need to strengthen the innovation ecosystem to allow AI researchers to externalize algorithms more easily to deliver public health products faster, specifically for health AI solutions intended for use by health care providers in low-resource clinical settings. Additionally, multiple common challenges across AI-based health products are surfacing as development progresses.

In community and primary health care settings, accessible, objective health information is vital for promptly identifying and directing resources to critically ill or at-risk patients.

Assessing the feasibility and performance of multimodal pulse oximeters

As part of the TIMCI project, PATH is leading an evaluation and operational research study to assess the feasibility and performance of multimodal pulse oximeters in primary healthcare settings across Kenya, Senegal, Tanzania, and Uttar Pradesh, India. This evaluation is essential to enhance the market for multimodal pulse oximeters by addressing uncertainties related to the validation and suitability of emerging clinical measurement tools derived from photoplethysmography (PPG). PPG is a measure of volumetric changes in blood circulation. From the PPG waveform blood oxygen saturation is derived, as well as parameters such as heart rate, respiratory rate, and total hemoglobin in simplified terms. PPG can be detected using existing sensors in a smartphone, such as the camera and flashlight, which opens many opportunities for use. Using digital biomarkers to better identify when and where health resources are needed is a cutting-edge AI opportunity for child health.

Using a hybrid type 2 research design that blends various diagnostic accuracy and implementation assessment methods, local primary care providers will evaluate the performance and practicality of these multimodal pulse oximeters. To ensure the seamless integration of these new health screening tools into the Integrated Management of Childhood Illness (IMCI) care process, the study engages end-users and key stakeholders through a collaborative co-design process.

Devices and manufacturers

Various devices are being evaluated for performance, feasibility, and sustainability in primary health care settings.

Scanbo (from Scanbo)
Neoguard (from Neopenda)
M-800 (from Biolight)
Model 3230 (from Nonin)
RadG+ (from Masimo)
Prototype noninvasive anemia application from Google Research

Devices and manufacturers are listed above from top left to bottom right.

Critical to successful product introduction and adoption is engaging users and key stakeholders early in the design process. A target product profile (TPP) was developed through a multi-phase process to define requirements of next generation multimodal PO devices. The TPP communicates requirements for optimal and minimum product attributes to align the global health community and manufacturers with the needs for the product class.

A human centered design workshop for next generation devices was conducted to engage country stakeholders, healthcare providers, and caregivers in defining user and health system constraints related to integrating smartphones and multimodal devices into IMCI practices. This work also informs manufacturers on desirable product attributes and improvements from a variety of perspectives to better ensure supply side and demand side alignment on future multimodal pulse oximetry products.

Finally, an open-access data repository of reference measurements collected in the diagnostic accuracy study is available upon request to catalyze PPG-derived, multimodal pulse oximters and smartphone-based clinical screening technology development. Improved access to key product development resources de-risks early R&D for needed health technologiesserving hard to reach populations, and allows developers to more easily validate new and available technologies to make the market more attractive for further industry investment.

Click here to inquire about accessing the data repository.

Unlocking AI’s Potential for Public Health through Smartphones and open access resources

AI is revolutionizing health care, offering solutions from disease diagnosis to drug discovery. Among the most promising frontiers in AI research is the development of smartphone-based health measurement technologies. These technologies harness smartphone sensors to gather data on various health indicators like heart rate, respiratory rate, temperature, and anemia. This data, along with AI models, can detect and monitor health conditions, and potentially predict which patients are at higher risk of severe illnesses.

However, translating AI models into accessible public health tools can be challenging, especially for developers who may not be familiar with commercializing health products or regulatory guidance for Software as a Medical Device. Recognizing the growing need to streamline the innovation process and help AI researchers bring public health tools to market in low and middle-income settings more efficiently, PATH is dedicated to strengthening the innovation ecosystem for AI-powered health technologies utilizing smartphones.

Our objectives include:

Forging partnerships to accelerate AI-driven health measurement technologies on existing smartphone platforms.
Facilitating knowledge sharing to address key barriers in deploying AI models.
Identifying synergies between AI model developers and health product creators.
Advocating for best practices concerning regulatory pathways and financing for AI-based health technologies.

Through PATH’s work to strengthen the innovation ecosystem for AI-powered health technologies utilizing smartphones, an AI for Health Toolkit was developed to share the learnings generated, including both the approach and the findings. This toolkit provides a broad overview and lessons learned from advancing the market for next-generation clinical measurement tools. Focus was given to common challenges impacting health AI developers and therefore included market and technology landscaping, partner evaluation and due diligence, regulatory review, evidence generation, and market analysis.

A key gap for this innovation space is the many similar but parallel initiatives to regulate software as a medical device as well as AI as a medical device. An additional resource created was an overview of the current AIaMD regulatory landscape and high-level guidance for product developers, governments, and global health donors to ensure all partners have a shared understanding of the current state of regulations and how improvements can be made in an ambiguous and constantly changing environment.

By sharing these learnings, PATH seeks to catalyze the availability of high-quality AI-assisted health tools that are appropriate for use in low-resource settings. Access to data, sharing insights and best practices, and fostering collaboration among diverse groups working on AI innovation can help solve common problems and accelerate progress, ensuring that solutions also reach underserved communities.