Smartphone App Accurate in Anemia Detection from Conjunctival Images

Gregory D. Jay, MD, PhD

Credit: Brown University

A new prospective, validation study deemed a smartphone application a useful, accurate tool for the non-invasive, real-time detection and classification of anemia using conjunctival images.¹

This study was the first to assess the clinical use of a smartphone-based app utilizing a novel 32-bit image processing method developed to predict anemia using digital images of the conjunctiva, with real-time local processing.

“Given the high prevalence of anemia and its associated morbidity and mortality, the development of an accurate and accessible anemia screening tool is essential to reduce the risk of health complications and alleviate the social and economic burdens associated with untreated anemia,” wrote the investigative team, led by Gregory D. Jay, MD, PhD, The Warren Alpert Medical School, Brown University.

Anemia impacts 22.8% of the global population, with young children, menstruating women, and pregnant or postpartum women particularly affected by low blood hemoglobin (Hb) concentration.² Clinical complications of anemia can include immune system dysfunction and slow cognitive and motor development—it is thus important to understand the socioeconomic determinants and create new tools to prioritize early diagnosis.

Standard tests for measuring anemia severity consist of the Complete Blood Count (CBC), but it requires significant resources and can take up to 4 hours to obtain the results.3 As a result, the collection of CBC can remain limited to geographic regions with adequate health infrastructure, despite anemia being disproportionately prevalent in areas lacking those resources.

Due to this, Jay and colleagues pointed to an unmet need for an accessible, non-invasive, point-of-care tool that allows rapid screening for anemia.¹ The performance of a smartphone imaging app that records RAW images of the palpebral conjunctiva and estimates Hb concentrations based on the tissue surface high hue ratio was measured for this analysis.

A convenience sample of patients (n = 435) who presented to the emergency department between June 2022 and February 2023 was used to obtain the images of bilateral conjunctivae using a dedicated smartphone. Participating patients were identified based on an electronic medical record alert that notified research assistants of an abnormal Hb level based on CBC.

Using the self-contained mobile app, investigators calculated high hue ratios from each image and correlated them to the laboratory-determined Hb (HBI). Conjunctival Hb concentration estimates (Hbc) were available instantly from the mobile app.

After excluding nine subjects due to only a singular recorded image, 426 subjects were left for the final analysis. The average age was 53 years and 55.2% were women.

Most patients (81.7%) demonstrated mild anemia by serology, with fewer subjects with severe (3.8%) and extreme (1.6%) anemia. However, the app showed the best accuracy in the extreme anemia category (< 5 g/dL), with agreement between HBc and HBl at 100%.

Overall, Jay and colleagues determined the accuracy of HBc was 75.4% (95% CI, 71.3 to 79.4) for all categories of anemia, being significantly lower for women (area under the curve [AUC], 0.74) than men (AUC, 0.79). The sensitivity and specificity of HBc for predicting anemia in men and women combined were 54.3 (95% CI, 46.9 - 61.8) and 89.7 (95% CI, 86.0 - 93.5), respectively.

Meanwhile, Bland-Altman plots were utilized to assess the agreement between HBc and HBl measures. The Bland-Altman plot analysis demonstrated a bias of 0.10 and limits of agreement (LOA) of (–4.73, 4.93 g/dL).

The investigative team also calculated receiver operating characteristic (ROC) curves to determine the impact of increasing transfusion thresholds on the sensitivity and specificity of HBc. The AUC for the restrictive transfusion threshold (<7 g/dL) was 0.92 (95% CI, 0.85 - 0.98), while the AUC for the liberal transfusion threshold was 0.90 (95% CI, 0.85 - 0.94).

Jay and colleagues indicated the smartphone app could demonstrate significant clinical utility across various health settings where the non-invasive screening of anemia would benefit care.

“As smartphones become ubiquitous, a non-invasive smartphone-based tool for anemia screening could serve as a pre-clinical and clinical tool that would allow for timely and inexpensive anemia screening for patients without access to blood testing,” Jay and colleagues.

References

1. _{Zhao L, Vidwans A, Bearnot CJ, et al. Prediction of anemia in real-time using a smartphone camera processing conjunctival images. PLoS One. 2024;19(5):e0302883. Published 2024 May 13. doi:10.1371/journal.pone.0302883}
2. _{Gardner W, Kassebaum N. Global, Regional, and National Prevalence of Anemia and Its Causes in 204 Countries and Territories, 1990–2019. Curr Dev Nutr. 2020;4(Suppl 2):830. Published 2020 May 29. doi:10.1093/cdn/nzaa053_035}
3. _{McLean E, Cogswell M, Egli I, Wojdyla D, de Benoist B. Worldwide prevalence of anaemia, WHO Vitamin and Mineral Nutrition Information System, 1993-2005. Public Health Nutr. 2009;12(4):444-454. doi:10.1017/S1368980008002401}