Fast and Automatic Ultrasound Simulation from CT Images

Towards improving breast cancer detection through multi-modal image generation

Ultrasound (US) imaging is real-time, less expensive, and more portable, compared to mammography, which makes it better suited for screening in resource-constrained settings and intra-operative imaging. However, US has lower spatial resolution and more artifacts compared to mammograms. This research aims to address these limitations by providing surgeons with mammogram-like image quality in real-time from US images. Previous approaches to US enhancement have discarded the artifacts created by interaction pattern between ultrasound and tissue by treating them as noise. By contrast, we recognize the value of the artifacts as wave interference patterns (WIP) that capture important tissue characteristics. In particular, we utilize the Stride software to numerically solve the forward model by generating US images from mammograms by solving wave-equations and add the high-frequency components separately to produce realistic US images. This forward generation itself is of clinical value because sometimes US acts as a complementary imaging modality to disambiguate cases that are difficult to diagnose using mammograms alone. Then, we train a generative adversarial network (GAN) for the obtaining mammogram-quality images from US. The resultant images have considerably more discernible details than the original US images. With further improvements, both forward and backward image generation can help simulate complementary modality on-the-fly to aid better breast cancer diagnosis in a cost-effective and real-time manner.

A novel cross-modal data augmentation method based on contrastive unpaired translation network for kidney segmentation in ultrasound imaging

BACKGROUND

Kidney ultrasound (US) image segmentation is one of the key steps in computer-aided diagnosis and treatment planning of kidney diseases. Recently, deep learning (DL) technology has demonstrated promising prospects in automatic kidney US segmentation. However, due to the poor quality, particularly the weak boundaries in kidney US imaging, obtaining accurate annotations for DL-based segmentation methods remain a challenging and time-consuming task. This issue can hinder the application of data-hungry deep learning methods.

PURPOSE

In this paper, we explore a novel cross-modal data augmentation method aimed at enhancing the performance of DL-based segmentation networks on the limited labeled kidney US dataset.

METHODS

In particular, we adopt a novel method based on contrastive unpaired translation network (CUT) to obtain simulated labeled kidney US images at a low cost from labeled abdomen computed tomography (CT) data and unlabeled kidney US images. To effectively improve the segmentation network performance, we propose an instance-weighting training strategy that simultaneously captures useful information from both the simulated and real labeled kidney US images. We trained our generative networks on a dataset comprising 4418 labeled CT slices and 4594 unlabeled US images. For segmentation network, we used a dataset consisting of 4594 simulated and 100 real kidney US images for training, 20 images for validation, and 169 real images for testing. We compared the performance of our method to several state-of-the-art approaches using the Wilcoxon signed-rank test, and applied the Bonferroni method for multiple comparison correction.

RESULTS

The experimental results show that we can synthesize accurate labeled kidney US images with a Fréchet inception distance of 52.52. Moreover, the proposed method achieves a segmentation accuracy of 0.9360 ± 0.0398 for U-Net on normal kidney US images, and 0.7719 ± 0.2449 on the abnormal dataset, as measured by the dice similarity coefficient. When compared to other training strategies, the proposed method demonstrated statistically significant superiority, with all p-values being less than 0.01.

CONCLUSIONS

The proposed method can effectively improve the accuracy and generalization ability of kidney US image segmentation models with limited annotated training data.

Principles for teaching sonography - current status

Since many young medical residents require sonographic skills early on during training, increased attention has been paid to including sonography classes in undergraduate medical education, among both professional societies and medical educators responsible for medical licensing exams. Medical schools worldwide have developed and implemented a variety of ultrasound teaching formats.This article addresses evidence-based solutions to crucial challenges in planning and implementing undergraduate sonography education. In order to achieve a sustainable increase in practical sonographic competence, we suggest small-group classes with sufficient individual hands-on scanning time for each student. We recommend concentrating on a circumscribed topic and teaching it thoroughly and practically rather than superficially outlining a broad subject area. Provided that peer teachers undergo adequate training, student peer teachers are not inferior to physicians as teachers, as far as student satisfaction, theoretical knowledge and practical skills acquisition are concerned. The assessment of acquired practical skills should consist of practical examinations, such as an objective structured clinical examination (OSCE) or a direct observation of procedural skills (DOPS). In contrast to using healthy volunteers as training models, simulation trainers allow the demonstration of pathological findings in authentic sonographic images, with the disadvantages of unrealistically easy image acquisition, as well as the lack of interaction with the patient.