Paper 53

Mobile app generated nail picture RGB data text files from very low birth weight (VLBW) infants at NICU bed: SAS code to assemble large volume of text files into analytical datasets for image analysis

September 23, 2025: 9:00 AM - 9:30 AM

Data Collection, Management, and Manipulation (DCMM), Executive Dining Room #3

Authors	Abstract
Neeta Shenvi	ABSTRACT This project presents the development and application of SAS macro code to process and integrate a large volume of RGB text files derived from fingernail and palm photos of preterm infants in the NICU bedside. Study nurse takes infant nail pictures in NICU bedside using a smart phone, identify a region of interest (a rectangle) on the picture. The phone app converts pixel intensities into numerical values for color (red, green, blue) and creates a text file. The text file from phone app is then uploaded to data coordinating center servers for further processing. The phone app generated over 1700 txt files from 89 infants (weekly longitudinal nail pictures for first 13 weeks of life during their stay in the NICU). We describe SAS 9.4 code to manage and merge these files to create analytic datasets. These datasets are further merged with infant longitudinal characteristics variables that are saved in a separate database. Our long term goal is to apply various image analysis algorithms on these analytic datasets to devise noninvasive methods to predict hemoglobin and detect anemia.

Neeta Shenvi

ABSTRACT This project presents the development and application of SAS macro code to process and integrate a large volume of RGB text files derived from fingernail and palm photos of preterm infants in the NICU bedside. Study nurse takes infant nail pictures in NICU bedside using a smart phone, identify a region of interest (a rectangle) on the picture. The phone app converts pixel intensities into numerical values for color (red, green, blue) and creates a text file. The text file from phone app is then uploaded to data coordinating center servers for further processing. The phone app generated over 1700 txt files from 89 infants (weekly longitudinal nail pictures for first 13 weeks of life during their stay in the NICU). We describe SAS 9.4 code to manage and merge these files to create analytic datasets. These datasets are further merged with infant longitudinal characteristics variables that are saved in a separate database. Our long term goal is to apply various image analysis algorithms on these analytic datasets to devise noninvasive methods to predict hemoglobin and detect anemia.