“Diagnóstico não invasivo em anemia: análise de foto da conjuntiva por smartphone e correlação com os níveis de hemoglobina”

Abstract

Anemia is a global health problem and mainly affects children and pregnant women, resulting in several health problems if it is not identified and treated early. Currently, the main diagnostic test used is the blood count, which measures the levels of hemoglobin in the patient's blood. To do so, it is necessary to collect blood using sharp materials, as well as use specific devices to analyze the collected material. Therefore, this research proposed to quantitatively analyze the color intensity in RGB of photographs of patients' conjunctiva, using a smartphone application, checking whether it is possible to establish a correlation between the color intensity detected in the image and the level of patient's hemoglobin measured by blood count and, using this method, optimize the process of diagnosing and screening for anemia. This is research with a quantitative and applied approach, whose objective is exploratory and the procedure adopted was field research. The sample consisted of 50 patients over 18 for whom a blood count test was requested. The photographs were analyzed using the RGB Color Detector mobile device application, which allows measuring the red, green, and blue channels (red, green, and blue - RGB), and the data was processed using the SPSS software. The Pearson correlation test was performed between Hemoglobin, Red Channel, Blue Channel, and Green Channel, in which only the Green channel correlated with the Red and Blue channels (p <0.001), both being weak correlations, concluding, therefore, that the correlation between the variables did not demonstrate potential as a diagnostic method as initially hypothesized, since a Pearson correlation above 0.5 (moderate to very strong) was expected. The performance of the new diagnostic test was evaluated using the ROC Curve (Receiver Operating Characteristic), which demonstrated that Canal Red alone has greater sensitivity than specificity. This variable (Canal Red) was able to detect 75% of cases in which there is anemia, and a lower capacity to correctly classify those who do not have anemia – 52% of cases, that is, the measurement of this couple showed potential as a non-invasive and In this aspect, it exceeded the sensitivity established as the project's initial target of 60%, showing that it is a promising method, despite its inherent limitations - small sampling, external interference and lack of more complex technologies for image analysis, as per proposed by most recent work in the area of interest. Finally, I conclude this work by reflecting on the production process of this research project as well as on my training in UFSCar medicine over the last six years.