Deeply-Supervised Networks With Threshold Loss for Cancer Detection in Automated Breast Ultrasound

Wang, Yi; Wang, Na; Xu, Min; Yu, Junxiong; Qin, Chenchen; Luo, Xiao; Yang, Xin; Wang, Tianfu; Liu, Lixian; Ni, Dong

doi:10.1109/tmi.2019.2936500

Cited by 126 publications

(41 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…e quantitative results suggested that the AC, Jaccard, and TP values of the LDL algorithm were greatly larger than those of the RA and ON algorithms, and the FP value of the LDL algorithm was much smaller in contrast to that of the RA and ON algorithms, showing statistically differences Journal of Healthcare Engineering (P < 0.05). Such results were similar to the findings of Wang et al [16], indicating that the LDL algorithm constructed in this study showed a good effect on tumor segmentation, improving the accuracy of tumor segmentation, so it was suitable for clinical promotion.…”

Section: Discussionsupporting

confidence: 91%

Ultrasound Image Features under Deep Learning in Breast Conservation Surgery for Breast Cancer

Zhang

Liu

et al. 2021

Journal of Healthcare Engineering

View full text Add to dashboard Cite

This study was to analyze the effect of the combined application of deep learning technology and ultrasound imaging on the effect of breast-conserving surgery for breast cancer. A deep label distribution learning (LDL) model was designed, and the semiautomatic segmentation algorithm based on the region growing and active contour technology (RA) and the segmentation model based on optimized nearest neighbors (ON) were introduced for comparison. The designed algorithm was applied to the breast-conserving surgery of breast cancer patients. According to the difference in intraoperative guidance methods, 102 female patients with early breast cancer were divided into three groups: 34 cases in W1 group (ultrasound guidance based on deep learning segmentation model), 34 cases in W2 group (ultrasound guidance), and 34 cases in W3 group (palpation guidance). The results revealed that the tumor area segmented by the LDL algorithm constructed in this study was closer to the real tumor area; the segmentation accuracy (AC), Jaccard, and true-positive (TP) values of the LDL algorithm were obviously greater than those of the RA and ON algorithms, while the false-positive (FP) value was significantly lower in contrast to the RA and ON algorithms, showing statistically observable differences ( P < 0.05); the actual resection volume of the patients in the W1 group was the closest to the ideal resection volume, which was much smaller in contrast to that of the patients in the W2 and W3 groups, showing statistical differences ( P < 0.05); the positive margins of the patients in the W1 group were statistically lower than those in the W2 and W3 groups ( P < 0.05). In addition, 1 patient in the W1 group was not satisfied with the cosmetic effect, 3 patients in the W2 group were not satisfied with the cosmetic effect, and 9 patients in the W3 group were not satisfied with the cosmetic effect. Finally, it was found that the ultrasound image based on the deep LDL model effectively improved the AC of tumor resection and negative margins, reduced the probability of normal tissue being removed, and improved the postoperative cosmetic effect of breast.

show abstract

Section: Discussionsupporting

confidence: 91%

Ultrasound Image Features under Deep Learning in Breast Conservation Surgery for Breast Cancer

Zhang

Liu

et al. 2021

Journal of Healthcare Engineering

View full text Add to dashboard Cite

show abstract

“…Their results achieved accuracy of 0.97% accuracy and AUC of 0.98% AUC. Another work is proposed in [40]. In the later work, the authors presented a method for breast cancer diagnosis utilising breast ultrasound pictures.…”

Section: Ultrasound Imagesmentioning

confidence: 99%

The Impact of Data processing and Ensemble on Breast Cancer Detection Using Deep Learning

Mohamed

Amer²,

Eldin

et al. 2022

Journal of Computing and Communication

View full text Add to dashboard Cite

According to the World Health Organization, cancer is the second leading cause of mortality. Breast cancer is the most prevalent cancer diagnosed in women around the world. Breast cancer diagnostics range from mammograms to CT scans and ultrasounds, but a biopsy is the only way to know for sure if the suspicious cells detected in the breast are cancerous or not. This paper's main contribution is multi-fold. First, it proposes a deep learning approach to detect breast cancer from biopsy microscopy images. Deep convolution nets of various types are used. Second, the paper examines the effects of different data preprocessing techniques on the performance of deep learning models. Third, the paper introduces an ensemble method for aggregating the best models in order to improve performance. The experimental results revealed that Densenet169, Resnet50, and Resnet101 are the three best models achieving accuracy scores of 62%, 68%, and 85%, respectively. without data preprocessing. With the help of data augmentation and segmentation, the accuracy of these models increased by 20%, 17%, and 6%, respectively. Additionally, the ensemble learning technique improves the accuracy of the models even further. The results show that the best accuracy achieved is 92.5%.

show abstract

“…The use of deep learning has proven to be very effective and reliable in revealing features, which are not evident, in images. Deep learning is currently widely used in the medical field for image classification and the detection of human diseases through computer-aided diagnosis [ 25 , 26 , 27 , 28 ]. Convolutional neural networks (CNNs) have demonstrated beneficial learning and useful feature extraction capabilities, and thus have been embraced by many researchers [ 29 ].…”

Section: Related Workmentioning

confidence: 99%

On the Use of Deep Learning for Imaging-Based COVID-19 Detection Using Chest X-rays

Okolo

Katsigiannis

Althobaiti

et al. 2021

Sensors

View full text Add to dashboard Cite

The global COVID-19 pandemic that started in 2019 and created major disruptions around the world demonstrated the imperative need for quick, inexpensive, accessible and reliable diagnostic methods that would allow the detection of infected individuals with minimal resources. Radiography, and more specifically, chest radiography, is a relatively inexpensive medical imaging modality that can potentially offer a solution for the diagnosis of COVID-19 cases. In this work, we examined eleven deep convolutional neural network architectures for the task of classifying chest X-ray images as belonging to healthy individuals, individuals with COVID-19 or individuals with viral pneumonia. All the examined networks are established architectures that have been proven to be efficient in image classification tasks, and we evaluated three different adjustments to modify the architectures for the task at hand by expanding them with additional layers. The proposed approaches were evaluated for all the examined architectures on a dataset with real chest X-ray images, reaching the highest classification accuracy of 98.04% and the highest F1-score of 98.22% for the best-performing setting.

show abstract

Deeply-Supervised Networks With Threshold Loss for Cancer Detection in Automated Breast Ultrasound

Cited by 126 publications

References 31 publications

Ultrasound Image Features under Deep Learning in Breast Conservation Surgery for Breast Cancer

Ultrasound Image Features under Deep Learning in Breast Conservation Surgery for Breast Cancer

The Impact of Data processing and Ensemble on Breast Cancer Detection Using Deep Learning

On the Use of Deep Learning for Imaging-Based COVID-19 Detection Using Chest X-rays

Contact Info

Product

Resources

About