End-to-end breast ultrasound lesions recognition with a deep learning approach

Yap, Moi Hoon; Goyal, Manu; Osman, Fatima; Ahmad, Ezak; Martí, Robert; Denton, Erika R. E.; Juette, Arne; Zwiggelaar, Reyer

doi:10.1117/12.2293498

Cited by 7 publications

(11 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…C. Experiments on the UDIAT Dataset 1) Comparison with Previous Methods: Comparisons are made with the recent approaches [18], [20], [21]. We first construct a system, comparable to the proposed method, by cascading the two methods [18], [20] because each method only conducts localization [18] or classification [20].…”

Section: B Experiments On the Snubh Dataset 1) Ablation Studymentioning

confidence: 99%

“…Each network constituting the whole system was separately trained using the UDIAT-DXLoc-Tr. We note that training the network in [18], [21] requires segmentation annotations, which are available only for the UDIAT dataset. The FCN-AlexNet shows the best performance in [18] in the absence of the FCN-VGG16, which showed the best performance in [45].…”

Section: B Experiments On the Snubh Dataset 1) Ablation Studymentioning

confidence: 99%

“…We thus use the FCN-VGG16 rather than the FCN-AlexNet in our experiment. The same modification on the network is applied for the method in [21]. For these methods [18], [20], [21], we use our own implementations because the authors did not make theirs publicly available.…”

Section: B Experiments On the Snubh Dataset 1) Ablation Studymentioning

confidence: 99%

“…Recently, more approaches based on deep learning are being proposed [17]- [21]. Many of these methods are limited to either localizing target objects of interest [17], [18] or classifying given RoIs into benign or malignant [19], [20], rather than conducting the two methods simultaneously.…”

mentioning

confidence: 99%

“…Many of these methods are limited to either localizing target objects of interest [17], [18] or classifying given RoIs into benign or malignant [19], [20], rather than conducting the two methods simultaneously. While the method recently proposed by Yap et al [21] performs both localization and classification simultaneously, it requires training data with lesion segmentation annotations since it is based on semantic image segmentation. This requirement may increase the burden of time and cost when compiling the training dataset.…”

mentioning

confidence: 99%

See 4 more Smart Citations

Joint Weakly and Semi-Supervised Deep Learning for Localization and Classification of Masses in Breast Ultrasound Images

Shin

Lee

Yun

et al. 2019

IEEE Trans. Med. Imaging

122

View full text Add to dashboard Cite

We propose a framework for localization and classification of masses in breast ultrasound (BUS) images. We have experimentally found that training convolutional neural network based mass detectors with large, weakly annotated datasets presents a non-trivial problem, while overfitting may occur with those trained with small, strongly annotated datasets. To overcome these problems, we use a weakly annotated dataset together with a smaller strongly annotated dataset in a hybrid manner. We propose a systematic weakly and semi-supervised training scenario with appropriate training loss selection. Experimental results show that the proposed method can successfully localize and classify masses with less annotation effort. The results trained with only 10 strongly annotated images along with weakly annotated images were comparable to results trained from 800 strongly annotated images, with the 95% confidence interval of difference -3.00%-5.00%, in terms of the correct localization (CorLoc) measure, which is the ratio of images with intersection over union with ground truth higher than 0.5. With the same number of strongly annotated images, additional weakly annotated images can be incorporated to give a 4.5% point increase in CorLoc, from 80.00% to 84.50% (with 95% confidence intervals 76.00%-83.75% and 81.00%-88.00%). The effects of different algorithmic details and varied amount of data are presented through ablative analysis.Index Terms-Breast ultrasound, convolutional neural networks, mass classification, mass localization, semisupervised learning, weakly supervised learning.

show abstract

Section: B Experiments On the Snubh Dataset 1) Ablation Studymentioning

confidence: 99%

Section: B Experiments On the Snubh Dataset 1) Ablation Studymentioning

confidence: 99%

Section: B Experiments On the Snubh Dataset 1) Ablation Studymentioning

confidence: 99%

mentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

Joint Weakly and Semi-Supervised Deep Learning for Localization and Classification of Masses in Breast Ultrasound Images

Shin

Lee

Yun

et al. 2019

IEEE Trans. Med. Imaging

122

View full text Add to dashboard Cite

show abstract

A Method for Localization and Classification of Breast Ultrasound Tumors

Zhu

Wang

2020

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Ultrasound instruments are suitable for large-scale examination of breast tumors, especially for women from Asian whose glands are dense. However, ultrasound images have the low contrast and resolution, blurred boundary and artifacts, which bring great difficulties to the interpretation of the junior doctor. However, traditional methods of breast ultrasound tumor recognition often use manually extracted features to gradually realize ROI region location and tumor classification with low accuracy, poor robustness and weak universality. Deep learning is limited to the location of tumor ROI region or the classification of a given tumor ROI region. In this paper, YOLOV3 algorithm is used for breast ultrasound tumor recognition, which could realize ROI localization and tumor classification at the same time. In addition, K-Means is optimized by K-Means++ and K-Mediods algorithm to generate anchor boxes of YOLOV3, and based on the Darknet-53 network structure of YOLOV3, ResNet and DenseNet are combined to design ResNet-DenseNet_Darknet-53. The proposed method is tested on the breast ultrasound tumor data set. Experiments show that the improved YOLOV3 algorithm shows better detection results on multiple evaluation indicators.

show abstract