Pengukuran Lingkar Lengan Atas Menggunakan Citra Digital : Pendekatan Top - Down Methods Panoptic Segmentation dengan Metode Mask R - CNN

Abstract

Measurement of Upper Arm Circumference (MUAC) is an essential indicator for evaluating nutritional status, especially in children and pregnant women. This study developed a UAC measurement system using digital images based on the Mask R-CNN method with a top-down panoptic segmentation approach. The model was implemented to detect the upper arm area in human body images and automatically calculate the UAC. Mask R-CNN facilitates more accurate segmentation of objects in complex images, particularly in the upper arm region. The training data was obtained from annotated images using the Roboflow platform, and the model was trained and evaluated to accurately detect and measure the UAC. Testing was conducted on 72 image samples, yielding a mean absolute error (MAE) of 2.31 cm between the system's measurements and manual measurements. Among the 72 samples, 21 individuals (29.2%) had a measurement difference of 0–1 cm, 20 individuals (27.8%) had a difference of 1–2 cm, and 31 individuals (43.1%) had a difference greater than 2 cm. Although there are still inconsistencies in the measurement results, the developed approach has potential for use in nutritional status assessments with higher levels of automation.