Learning based automatic head detection and measurement from fetal ultrasound images via prior knowledge and imaging parameters

Abstract

A novel learning based automatic method is proposed to detect the fetal head for the measurement of head circumference from ultrasound images. We first exploit the AdaBoost learning method to train the classifier on Haar-like features and then, for the first time, we propose to use prior knowledge and online imaging parameters to guide the sliding window based head detection from ultrasound images. This approach can significantly improve both detection rate and speed. The boundary of the head in the localized region is further detected using a local phase based method, which is insensitive to speckle noises and intensity changes in ultrasound images. Finally iterative randomized Hough transform (IRHT) is employed to determine an ellipse on the head contour. Experiments performed on 675 images (500 for classifier training and 175 for measurement) showed that mean-signeddifference between automatic and manual measurements is 2.86 mm (1.6%). The statistical analysis further indicated that there was no significant difference between these two measurements. These results demonstrated the proposed fully automatic framework can be used as a consistent and accurate tool in clinical practice.

A review of image processing methods for fetal head and brain analysis in ultrasound images

“…Indeed, phase-based methods showed high accuracy and good response to speckle noise in US images, being useful to improve edge detection. Thus, Ni et al [50] and Li J et al [51] proposed to use this type of method to detect skull edges. Both methods performed detection of head's ROI before applying edge detection, using Adaboost classifier [50] and RF [51] .…”

STNet: shape and texture joint learning through two-stream network for knowledge-guided image recognition

“…Han et al ( 2020 ) proposed an ensemble learning method for panoramic radiographs recognition based on the characteristics of each stage of tooth growth. Ni et al ( 2013 ) proposed a novel learning-based automatic method to detect the fetal head for the measurement of head circumference from ultrasound images and used prior knowledge and online imaging parameters to guide the sliding window-based head detection. Pan et al ( 2022 ) proposed a two-stage network with prior knowledge guidance for medullary thyroid carcinoma recognition in ultrasound images.…”

Automated localization and segmentation techniques for B-mode ultrasound images: A review

“…The fetal femur can tend to lack internal texture but usually presents strong edges in most of the contour except for in the extremities [220]. Segmentation methods for the fetal head and femur detection vary greatly, including the Hough transform, morphologic operators, and active contours, but recent research has focused a lot on the calculation of features and subsequent classification, including Haar-like features [61], texton cues [221], and shape information of pixel groups and local statistics [67]. These methods have the advantage of being able to incorporate many different types of image features and prior knowledge, making the final segmentation more robust.…”