Evaluation of deep convolutional neural networks for automatic classification of common maternal fetal ultrasound planes

Abstract

The goal of this study was to evaluate the maturity of current Deep Learning classification techniques for their application in a real maternal-fetal clinical environment. A large dataset of routinely acquired maternal-fetal screening ultrasound images (which will be made publicly available) was collected from two different hospitals by several operators and ultrasound machines. All images were manually labeled by an expert maternal fetal clinician. Images were divided into 6 classes: four of the most widely used fetal anatomical planes (Abdomen, Brain, Femur and Thorax), the mother's cervix (widely used for prematurity screening) and a general category to include any other less common image plane. fetal brain images were further categorized into the 3 most common fetal brain planes (Trans-thalamic, Trans-cerebellum, Trans-ventricular) to judge fine grain categorization performance. The final dataset is comprised of over 12,400 images from 1,792 patients, making it the largest ultrasound dataset to date. We then evaluated a wide variety of state-of-the-art deep convolutional neural networks on this dataset and analyzed results in depth, comparing the computational models to research technicians, which are the ones currently performing the task daily. Results indicate for the first time that computational models have similar performance compared to humans when classifying common planes in human fetal examination. However, the dataset leaves the door open on future research to further improve results, especially on fine-grained plane categorization.

Evaluation of transfer learning in deep convolutional neural network models for cardiac short axis slice classification

“…rather than medical images. Burgos-Artizzu et al evaluated 19 deep CNN models including VGG16 for the classification of fetal ultrasound images and reported that DenseNet-169 was the best performing model in terms of top-1 error 34 . This suggests that the transferability of ImageNet models is application-dependent and can vary depending on the difference between source domain and target domain.…”

Towards Clinical Application of Artificial Intelligence in Ultrasound Imaging

“…Furthermore, when labeling a huge amount of US images for AI-based image processing, it is necessary to classify the acquired US images and to assess whether the image quality thereof is suitable for the input data. Burgos-Artizzu et al evaluated a wide variety of CNNs for the automatic classification of a large dataset containing over 12,400 images from 1792 patients that were routinely acquired during maternal-fetal US screening [125]. An automatic recognition method using deep learning for the fetal facial standard planes, including the axial, coronal, and sagittal planes was reported [126].…”

Reverse Scan Conversion and Efficient Deep Learning Network Architecture for Ultrasound Imaging on a Mobile Device

“…Therefore, the ultrasound image reconstruction for the next frame is delayed. We call this the frame synchronous classification (FSC) structure [27], which has the disadvantage of delaying t image as much as the classification processing time (t CP ). To overcome this disadvantage of the FSC structure, we propose a structure that can implement ultrasound image reconstruction and a classification network in real time on a mobile device, as shown in Figure 5c.…”