Densely Deep Supervised Networks with Threshold Loss for Cancer Detection in Automated Breast Ultrasound

Wang, Na; Bian, Cheng; Wang, Yi; Xu, Min; Qin, Chenchen; Yang, Xin; Wang, Tianfu; Liu, Lixian; Shen, Dinggang; Ni, Dong

doi:10.1007/978-3-030-00937-3_73

Cited by 25 publications

(20 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Ultrasound imaging is a standard modality for many diagnostic and monitoring purposes, and there has been significant research into developing automatic methods for segmentation of ultrasound images [5,11]. U-Net [7] for instance has been shown to be a fast and precise solution for medical image segmentation, and has successfully been adapted to segment ultrasound images too [10,1,8,12]. In this study, we investigate the effect of fine-tuning different layers of a U-Net network for the application of ultrasound image segmentation.…”

Section: Introductionmentioning

confidence: 99%

Fine Tuning U-Net for Ultrasound Image Segmentation: Which Layers?

Amiri

Brooks

Rivaz

2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Fine-tuning a network which has been trained on a large dataset is an alternative to full training in order to overcome the problem of scarce and expensive data in medical applications. While the shallow layers of the network are usually kept unchanged, deeper layers are modified according to the new dataset. This approach may not work for ultrasound images due to their drastically different appearance. In this study, we investigated the effect of fine-tuning different layers of a U-Net which was trained on segmentation of natural images in breast ultrasound image segmentation. Tuning the contracting part and fixing the expanding part resulted in substantially better results compared to fixing the contracting part and tuning the expanding part. Furthermore, we showed that starting to finetune the U-Net from the shallow layers and gradually including more layers will lead to a better performance compared to fine-tuning the network from the deep layers moving back to shallow layers. We did not observe the same results on segmentation of X-ray images, which have different salient features compared to ultrasound, it may therefore be more appropriate to fine-tune the shallow layers rather than deep layers. Shallow layers learn lower level features (including speckle pattern, and probably the noise and artifact properties) which are critical in automatic segmentation in this modality.

show abstract

Section: Introductionmentioning

confidence: 99%

Fine Tuning U-Net for Ultrasound Image Segmentation: Which Layers?

Amiri

Brooks

Rivaz

2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…We further extensively compared our network with the cutting-edge networks, including 3D FCN, 29 Vanilla Unet, 19 Residual Unet, 30 and our previous breast cancer detection (BCD) model. 17 Figure 8 illustrates the free-response receiver operating characteristic curves (FROCs) of the proposed network and compared models. Table I reports the numerical results of sensitivities, FPs per ABUS volume, IoU, and Cen-Dis for different networks.…”

Section: D Detection Performancementioning

confidence: 99%

“…Experimental results show the asymmetric loss outperformed CE loss and DSC loss with respect to detection sensitivity, IoU, and CenDis, which demonstrates the asymmetric loss contributed to the cancer detection. Figure 10 further shows the volume distribution of all cancerous regions, and their corresponding detection sensitivities achieved by BCD model 17 and our proposed network, respectively. It is shown in Fig.…”

Section: D Detection Performancementioning

confidence: 99%

3D Inception U‐net with Asymmetric Loss for Cancer Detection in Automated Breast Ultrasound

et al. 2020

Self Cite

View full text Add to dashboard Cite

Purpose Breast cancer is the most common cancer and the leading cause of cancer‐related deaths for women all over the world. Recently, automated breast ultrasound (ABUS) has become a new and promising screening modality for whole breast examination. However, reviewing volumetric ABUS is time‐consuming and lesions could be missed during the examination. Therefore, computer‐aided cancer detection in ABUS volume is extremely expected to help clinician for the breast cancer screening. Methods We develop a novel end‐to‐end 3D convolutional network for automated cancer detection in ABUS volume, in order to accelerate reviewing and meanwhile to provide high detection sensitivity with low false positives (FPs). Specifically, an efficient 3D Inception Unet‐style architecture with fusion deep supervision mechanism is proposed to attain decent detection performance. In addition, a novel asymmetric loss is designed to help the network balancing false positive and false negative regions, thus improving detection sensitivity for small cancerous lesions. Results The efficacy of our network was extensively validated on a dataset including 196 patients with 661 cancer regions. Our network obtained a detection sensitivity of 95.1% with 3.0 FPs per ABUS volume. Furthermore, the average inference time of the network was 0.1 second per volume, which largely shortens the conventional reviewing time. Conclusions The proposed network provides efficient and accurate cancer detection scheme using ABUS volume, and may assist clinicians for more efficient breast cancer screening.

show abstract

“…To the best of our knowledge, only two methods have been developed for lesion detection in ABUS images. As the first attempt, Wang et al (2018) utilized a 3-D U-Net basically introduced in (Ronneberger et al 2015) as their backbone architecture and improved its performance by some modifications. Due to limited number of ABUS training samples as well as avoiding over-fitting issue, they used a pre-trained 3-D convolutional network (Tan et al 2015) and fine-tuned the hyper-parameters of the network.…”

Section: Deep Learning Based Approachesmentioning

confidence: 99%

Computer aided detection in automated 3-D breast ultrasound images: a survey

et al. 2019

View full text Add to dashboard Cite

Nowadays, breast cancer is the leading cause of cancer death for women all over the world. Since the reason of breast cancer is unknown, early detection of the disease plays an important role in cancer control, saving lives and reducing costs. Among different modalities, automated 3-D breast ultrasound (3-D ABUS) is a new and effective imaging modality which has attracted a lot of interest as an adjunct to mammography for women with dense breasts. However, reading ABUS images is time consuming for radiologists and subtle abnormalities may be overlooked. Hence, computer aided detection (CADe) systems can be utilized as a second interpreter to assist radiologists to increase their screening speed and sensitivity. In this paper, a general architecture representing different CADe systems for ABUS images is introduced and the approaches for implementation of each block are categorized and reviewed. In addition, the limitations of these systems are discussed and their performance in terms of sensitivity and number of false positives per volume are compared.

show abstract

Densely Deep Supervised Networks with Threshold Loss for Cancer Detection in Automated Breast Ultrasound

Cited by 25 publications

References 10 publications

Fine Tuning U-Net for Ultrasound Image Segmentation: Which Layers?

Fine Tuning U-Net for Ultrasound Image Segmentation: Which Layers?

3D Inception U‐net with Asymmetric Loss for Cancer Detection in Automated Breast Ultrasound

Computer aided detection in automated 3-D breast ultrasound images: a survey

Contact Info

Product

Resources

About