Shining Light on Dark Skin: Pulse Oximetry Correction Models

Matos, João; Struja, Tristan; Gallifant, Jack; Charpignon, Marie-Laure; Cardoso, Jaime S.; Celi, Leo Anthony

doi:10.1109/enbeng58165.2023.10175316

Cited by 6 publications

(5 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This dataset comprises ICU patients – a population that was able to reach the hospital to receive critical care in the first place – introducing a potential selection bias that should be noted in any study using this dataset. Algorithms that are trained on MIMIC-IV data should target a population that is adequately represented in the dataset, otherwise both dataset and subpopulation shifts may occur [49, 50].…”

Section: Example Case Studymentioning

confidence: 99%

“…The above shifts were flagged by our diagram, and can be shown by the measured SMD for these two categories. These changes could result for a number of reasons, which include certain demographics being less likely to receive an ABG measurement in the hospital [50]. The participant flow diagram allows us to see the disproportionate dropout and to inform the study and the findings, which could lead to an investigator adjusting the study as a whole to lead to more inclusion or reporting this as a key limitation depending on the situation.…”

Section: Example Case Studymentioning

confidence: 99%

“…Specifically, we discuss the sample selection process for training such a model using real data from the MIMIC-IV dataset [49]. However, the creation of such models can be limited by a disproportionate absence of ABG measurements, resulting in a higher rate of exclusion of Black patients [50].…”

Section: Example Case Studymentioning

confidence: 99%

See 2 more Smart Citations

Participant Flow Diagrams for Health Equity in AI Research

Ellen,

Matos,

Viola

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Biases in sample creation can arise at any study phase, including initial patient recruitment, exclusion criteria, input-level exclusion and outcome-level exclusion, and often reflect the underrepresentation or exclusion of demographic groups historically disadvantaged in medical research. The use of non-representative samples to construct clinical algorithms in artificial intelligence (AI) and machine learning (ML) applications may further amplify this selection bias. Building on the “Data Cards” initiative for transparency in AI research, we advocate for the addition of a detailed participant flow diagram for AI studies, emphasizing the need to detail excluded participant demographic characteristics at every study phase. This tracking of excluded participants enhances understanding of potential algorithmic biases before their clinical implementation, and thus deserves to be detailed in any medical AI study. We include both a model for this flow diagram as well as a brief case study explaining how it could be implemented in practice. Through standardized reporting of participant flow diagrams, we can better gauge the potential inequity embedded in AI applications, facilitating more reliable and equitable clinical algorithms.

show abstract

Section: Example Case Studymentioning

confidence: 99%

Section: Example Case Studymentioning

confidence: 99%

See 1 more Smart Citation

Participant Flow Diagrams for Health Equity in AI Research

Ellen,

Matos,

Viola

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Our work stands as a necessary and fundamental milestone to any machine learning (ML) 8 or advanced data analysis of oxygen readings that can be performed to produce actionable insights. Data curation is a critical step, especially considering the volume of data in our base datasets; e.g., MIMIC-III v.1.4 alone contains over 58,000 hospital admissions from approximately 38,600 adults, resulting in 6.2GB of data.…”

Section: Background and Summarymentioning

confidence: 99%

BOLD: Blood-gas and Oximetry Linked Dataset – Open Source Research

Matos,

Struja,

Gallifant

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Pulse oximeters measure peripheral arterial oxygen saturation (SpO2) noninvasively, while the gold standard (SaO2) involves arterial blood gas measurement. There are known racial and ethnic disparities in their performance. BOLD is a new comprehensive dataset that aims to underscore the importance of addressing biases in pulse oximetry accuracy, which disproportionately affect darker-skinned patients.The dataset was created by harmonizing three Electronic Health Record databases (MIMIC-III, MIMIC-IV, eICU-CRD) comprising Intensive Care Unit stays of US patients. Paired SpO2and SaO2measurements were time-aligned and combined with various other sociodemographic and parameters to provide a detailed representation of each patient. BOLD includes 49,099 paired measurements, within a 5-minute window and with oxygen saturation levels between 70-100%. Minority racial and ethnic groups account for ∼25% of the data – a proportion seldom achieved in previous studies. The codebase is publicly available.Given the prevalent use of pulse oximeters in the hospital and at home, we hope that BOLD will be leveraged to develop debiasing algorithms that can result in more equitable healthcare solutions.

show abstract

“…The work of Matos et al was also devoted to the study of the accuracy of pulse oximeters in a population with a darker skin tone [102]. The results of their single-center study suggest that SpO2 correction can be achieved with machine learning.…”

mentioning

confidence: 99%

Patent and Bibliometric Analysis of the Scientific Landscape of the Use of Pulse Oximeters and Their Prospects in the Field of Digital Medicine

Litvinova,

Hammerle,

Stoyanov

et al. 2023

Healthcare

View full text Add to dashboard Cite

This study conducted a comprehensive patent and bibliometric analysis to elucidate the evolving scientific landscape surrounding the development and application of pulse oximeters, including in the field of digital medicine. Utilizing data from the Lens database for the period of 2000–2023, we identified the United States, China, the Republic of Korea, Japan, Canada, Australia, Taiwan, and the United Kingdom as the predominant countries in patent issuance for pulse oximeter technology. Our bibliometric analysis revealed a consistent temporal trend in both the volume of publications and citations, underscoring the growing importance of pulse oximeters in digitally-enabled medical practice. Using the VOSviewer software(version 1.6.18), we discerned six primary research clusters: (1) measurement accuracy; (2) integration with the Internet of Things; (3) applicability across diverse pathologies; (4) telemedicine and mobile applications; (5) artificial intelligence and deep learning; and (6) utilization in anesthesiology, resuscitation, and intensive care departments. The findings of this study indicate the prospects for leveraging digital technologies in the use of pulse oximetry in various fields of medicine, with implications for advancing the understanding, diagnosis, prevention, and treatment of cardio-respiratory pathologies. The conducted patent and bibliometric analysis allowed the identification of technical solutions to reduce the risks associated with pulse oximetry: improving precision and validity, technically improved clinical diagnostic use, and the use of machine learning.

show abstract

Shining Light on Dark Skin: Pulse Oximetry Correction Models

Cited by 6 publications

References 6 publications

Participant Flow Diagrams for Health Equity in AI Research

Participant Flow Diagrams for Health Equity in AI Research

BOLD: Blood-gas and Oximetry Linked Dataset – Open Source Research

Patent and Bibliometric Analysis of the Scientific Landscape of the Use of Pulse Oximeters and Their Prospects in the Field of Digital Medicine

Contact Info

Product

Resources

About