Robust border enhancement and detection for measurement of fetal nuchal translucency in ultrasound images

Artificial Intelligence in Prenatal Ultrasound Diagnosis

The application of artificial intelligence (AI) technology to medical imaging has resulted in great breakthroughs. Given the unique position of ultrasound (US) in prenatal screening, the research on AI in prenatal US has practical significance with its application to prenatal US diagnosis improving work efficiency, providing quantitative assessments, standardizing measurements, improving diagnostic accuracy, and automating image quality control. This review provides an overview of recent studies that have applied AI technology to prenatal US diagnosis and explains the challenges encountered in these applications.

Automatic measurement of fetal Nuchal translucency from three-dimensional ultrasound data

The Nuchal translucency (NT), which is the collection of fluid at the back of the fetal neck, is related to chromosomal defects and early cardiac failure in first trimester of pregnancy. In clinic, the thickness of NT is used as an important marker in prenatal screening, and is manually measured by sonographers in the mid-sagittal plane. In this paper, an automatic method based on dynamic programming is proposed to detect the thickness and area of NT in the mid-sagittal plane. Furthermore, the volume of NT in the whole three-dimensional ultrasound data is also measured. A novel cost function for dynamic programming is proposed and results in higher accuracy of NT border detection. As the nuchal translucency is the collection fluid part, higher dimensional markers of NT possess more potential to represent the amount of the fluid.

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

B-mode ultrasound imaging is used extensively in medicine. Hence, there is a need to have efficient segmentation tools to aid in computer-aided diagnosis, image-guided interventions, and therapy. This paper presents a comprehensive review on automated localization and segmentation techniques for B-mode ultrasound images. The paper first describes the general characteristics of B-mode ultrasound images. Then insight on the localization and segmentation of tissues is provided, both in the case in which the organ/tissue localization provides the final segmentation and in the case in which a two-step segmentation process is needed, due to the desired boundaries being too fine to locate from within the entire ultrasound frame. Subsequenly, examples of some main techniques found in literature are shown, including but not limited to shape priors, superpixel and classification, local pixel statistics, active contours, edge-tracking, dynamic programming, and data mining. Ten selected applications (abdomen/kidney, breast, cardiology, thyroid, liver, vascular, musculoskeletal, obstetrics, gynecology, prostate) are then investigated in depth, and the performances of a few specific applications are compared. In conclusion, future perspectives for B-mode based segmentation, such as the integration of RF information, the employment of higher frequency probes when possible, the focus on completely automatic algorithms, and the increase in available data are discussed.

Automatic detection and measurement of nuchal translucency

In this paper we propose a new methodology to support the physician both to identify automatically the nuchal region and to obtain a correct thickness measurement of the nuchal translucency. The thickness of the nuchal translucency is one of the main markers for screening of chromosomal defects such as trisomy 13, 18 and 21. Its measurement is performed during ultrasound scanning in the first trimester of pregnancy. The proposed methodology is mainly based on wavelet and multi resolution analysis. The performance of our method was analysed on 382 random frames, representing mid-sagittal sections, uniformly extracted from real clinical ultrasound videos of 12 patients. According to the ground-truth provided by an expert physician, we obtained a true positive rate equal to 99.95% with respect to the nuchal region detection and about 64% of measurements present an error equal to 1 pixel (which corresponds to 0.1mm), respectively.

Automatic Detection of Standard Sagittal Plane in the First Trimester of Pregnancy Using 3-D Ultrasound Data

Fetal nuchal translucency (NT) thickness is one of the most important parameters in prenatal screening. Locating the mid-sagittal plane is one of the key points to measure NT. In this paper, an automatic method for the sagittal plane detection using 3-D ultrasound data is proposed. To avoid unnecessary massive searching and the corresponding huge computation load, a model is proposed to turn the sagittal plane detection problem into a symmetry plane and axis searching problem. The deep belief network (DBN) and a modified circle detection method provide prior knowledge for the searching. The experiments show that in most cases, the result plane has small distance error and angle error at the same time-88.6% of the result planes have a distance error less than 4 mm and 71.0% have angle error less than 20°.

Image Settings Affecting Nuchal Translucency Measurement Using Volume NT Software

PURPOSE

To evaluate the effects of the deviation from the mid-sagittal plane, fetal image size, tissue harmonic imaging (THI), and speckle reduction filter (SRF) on the measurement of the nuchal translucency (NT) thickness using Volume NT software.

MATERIALS AND METHODS

In 79 pregnant women, NT was measured using Volume NT. Firstly, the three-dimensional volumes were categorized based on the angle of deviation in 10° intervals from the mid-sagittal plane. Secondly, the operator downsized the fetal image to less than 50% of the screen (Method A) and by magnifying the image (Method B). Next, the image was magnified until the fetal head and thorax occupied 75% of the screen, and the NT was measured (Method C). Lastly, NT values were acquired with THI and SRF functions on, with each function alternately on, and with both functions off.

RESULTS

The mean differences in NT measurements were -0.09 mm (p<0.01) between two-dimensional (2D) and a deviation of 31-40° and -0.10 mm (p<0.01) between 2D and 41-50°. The intraclass correlation coefficients (ICC) for 2D-NT and NT according to image size were 0.858, 0.923, and 0.928 for methods A, B, and C, respectively. The ICC for 2D-NT and NT with respect to the THI and SRF were 0.786, 0.761, 0.740, and 0.731 with both functions on, THI only, SRF only, and with both functions off, respectively.

CONCLUSION

NT measurements made using Volume NT are affected by angle deviation from the mid-sagittal plane and fetal image size. Additionally, the highest correlation with 2D-NT was achieved when THI and SRF functions were used.

Effect of image settings on nuchal translucency thickness measurement by a semi-automated system

OBJECTIVE

To investigate whether pre- and post-processing image settings affect NT measurements made by a semi-automatic method (SAM).

METHODS

Different image settings (e.g. gain) were either adjusted one at a time on images that had been obtained during fetal NT scans (post-processing, n = 66), or adjusted one at a time during live scanning and image acquisition of the adult posterior tibial artery (pre-processing group, n = 91). The NT and luminal diameter of the posterior tibial artery, respectively, were measured by SAM on all original and adjusted images.

RESULTS

Alteration of the image settings resulted in a statistically significant effect on the measurements taken by SAM, with an average pair difference ranging from 0.001 mm to 0.139 mm. Most of the differences were small and therefore the clinical impact would be negligible. The pair differences were greatest with a very high contrast setting, or without tissue harmonic imaging (THI); the paired difference in measurement in those with vs those without THI was more than 0.1 mm in over 40% of cases.

CONCLUSIONS

Measurements made by SAM are affected by image settings.

Semi-automated adjusted measurement of nuchal translucency: feasibility and reproducibility

OBJECTIVE

The variability of nuchal translucency thickness (NT) measurements in the first trimester appears to be associated in part with caliper placement. Methods for obtaining semi-automated adjusted measurements (SAAMs) can provide several NT values (maximum, minimum, mean and median) automatically within a manually set frame in the zone of interest. This study sought to assess the feasibility and reproducibility of these SAAM-NTs.

METHODS

Three readers, two experts and one less experienced, examined archive images of 160 patients and obtained SAAM-NTs from them, on two separate occasions. The intra- and interobserver reproducibility were assessed by calculating the intraclass correlation coefficients (ICCs) for maximum, mean and median SAAM-NTs, and Bland-Altman plots were constructed.

RESULTS

SAAM-NTs were technically feasible for all 160 images. The range of ICCs for intraobserver reproducibility was 0.76-0.93 for mean SAAM-NT, 0.76-0.95 for median SAAM-NT and 0.74-0.95 for maximum SAAM-NT. Interobserver ICCs were 0.85, 0.85 and 0.84 for mean SAAM-NT, median SAAM-NT and maximum SAAM-NT, respectively. There were no significant differences for intra- and interobserver reproducibility of median, mean and maximum SAAM-NTs.

CONCLUSIONS

SAAM-NT is feasible with a high level of intra- and interobserver reproducibility. This easy-to-use method has the potential to simplify screening during the first trimester. It should be evaluated further and compared with the manual measurement method.

Semi-automated system for measurement of nuchal translucency thickness

OBJECTIVE

To estimate intersonographer and intrasonographer variance components of fetal nuchal translucency (NT) thickness measurement using the traditional manual approach and a new semi-automated system.

METHODS

A semi-automated method was developed for measurement of the NT. In this method, the operator places an adjustable box over the relevant area at the back of the fetal neck. The system draws a line through the center of the nuchal membrane and another line at the edge of the soft tissue overlying the cervical spine. The system then identifies the largest vertical distance between the two lines. The images of 12 fetuses at 11-13 weeks of gestation satisfying the guidelines of The Fetal Medicine Foundation for measurement of NT were selected. They were exported in DICOM format from the ultrasound system, and four versions of each image were stored under different names. The resulting 48 images were presented in random order for electronic assessment. A total of 20 sonographers measured the NT in each set of 48 pictures, twice using the semi-automated system and twice using the manual system, according to a randomized block design. Within- and between-operator variance components were estimated. Relative biases were assessed by comparing the means from the two methods.

RESULTS

The estimated between-operator SD using the semi-automated method was 0.0149 mm compared with 0.109 mm for the manual method. The respective within-operator SD values were 0.05 mm and 0.126 mm. The intraclass correlation coefficients for different sonographers measuring the same images were 0.98 and 0.85 for the semi-automated method and the manual method, respectively.

CONCLUSION

The measurement of fetal NT is more reliable when a semi-automatic approach is used rather than the traditional manual method.

An informative probability model enhancing real time echobiometry to improve fetal weight estimation accuracy

A multinormal probability model is proposed to correct human errors in fetal echobiometry and improve the estimation of fetal weight (EFW). Model parameters were designed to depend on major pregnancy data and were estimated through feed-forward artificial neural networks (ANNs). Data from 4075 women in labour were used for training and testing ANNs. The model was implemented numerically to provide EFW together with probabilities of congruence among measured echobiometric parameters. It enabled ultrasound measurement errors to be real-time checked and corrected interactively. The software was useful for training medical staff and standardizing measurement procedures. It provided multiple statistical data on fetal morphometry and aid for clinical decisions. A clinical protocol for testing the system ability to detect measurement errors was conducted with 61 women in the last week of pregnancy. It led to decisive improvements in EFW accuracy.