Prediction of CBC Hemoglobin Levels for Preterm Infants: Evaluation of Anatomic Regions for Smartphone Photos Público

Wang, Xinzhu (Spring 2022)

Permanent URL: https://etd.library.emory.edu/concern/etds/5138jg077?locale=pt-BR
Published

Abstract

Anemia is a condition in which the body’s red blood cell count is lower than normal and become unable to provide enough oxygen to the tissues. If untreated, it leads to a delay in brain maturation, tissue hypoxia, and growth impairment. The routine techniques to detect anemia are blood laboratory tests such as complete blood count (CBC). These tests require frequent blood draws, which may cause phlebotomy-induced complications or exacerbate the existing anemia.

Non-invasive technologies are needed for preterm infants to quantitatively estimate their hemoglobin levels and thereby monitoring the anemia status. Inspired by a recent study using smartphone photos of adults’ fingernail beds for non-invasive detection of anemia, we aim to explore whether a non-invasive method that is similar to this smartphone image analysis algorithm can be developed for the preterm infants. The goal of this study is to investigate several anatomic regions and to determine smartphone photos of which anatomic region (fingernail, palm or toenail) can most accurately reflect the preterm infants’ CBC Hgb levels. Linear regression models are fitted to predict the hemoglobin outcomes for all anatomic regions as well as individually for each anatomic region. Mean squared errors and mean absolute errors for all models across all testing data are calculated to compare the prediction performance.

The anatomic region is significantly associated with the CBC Hgb levels, indicating that different anatomic regions have different intercepts for the regression equation. The mean predicted Hgb levels at palm region is significantly different from the mean predicted Hgb levels at fingernail region (p-value = 0.028). The regression model for all anatomic regions has a mean absolute error of 1.26 and mean squared error of 2.53 based on the testing data that includes all anatomic regions. The higher MSE and absolute errors are found with palm and fingernail testing data while the error is slightly smaller for the toenail testing data.

Depending on the anatomic regions, the predictive equations can be different. The regression equation that includes all anatomic regions appear to have similar prediction errors across all anatomic regions, but the toenail testing data may provide slightly smaller prediction errors compared with the fingernail testing data and palm testing data.

Table of Contents

1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.1 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.2 Statistical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.1 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.2 Prediction Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

Figures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Appendix. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 

About this Master's Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
School
Department
Degree
Submission
Language
  • English
Research Field
Palavra-chave
Committee Chair / Thesis Advisor
Committee Members
Última modificação

Primary PDF

Supplemental Files