®®®® SIIA Público

Título del libro:
Título del capítulo: Deep Learning Applied to Automatic Fetometry

Autores UNAM:
FERNANDO ARAMBULA COSIO;
Autores externos:
Idioma:
Año de publicación:
2025
Palabras clave:

Fetal ultrasound; Fetometry; Convolutional neural networks

Resumen:

In this work we report preliminary results of a, deep learning based, system designed for remote assessment of fetal growth, through automatic plane selection, and fetometry measurements in fetal ultrasound images. The system is designed to measure automatically: the circumference and biparietal diameter of the fetal head; the circumference of the abdomen; and the length of the femur. We have trained different convolutional neural networks (CNNs) for region detection and region segmentation. We report our preliminary results of region detection for head, abdomen and femur, using a YOLO V2 CNN, on 3 tests sets of 50 images each. The mean errors produced in the location, and size, of the bounding, boxes by the YOLO V2 are: 17 and 13 pixels for the top left coordinates (x, y); 24 and 19 pixels for the width and height of the bounding boxes. We also report our preliminary results of femur segmentation using a u-net CNN, with a 0.69 Dice coefficient on 10 test images.

Entidades citadas de la UNAM:
Fuente:



ISBN:
9783031821226
Editorial:
Springer Science and Business Media Deutschland GmbH

Tipo de producto:
Proceedings Paper



DOI:
10.1007/978-3-031-82123-3_31