Computerized detection and classification of cancer on breast ultrasound1

doi:10.1016/S1076-6332(03)00723-2

Academic Radiology

Volume 11, Issue 5, May 2004, Pages 526-535

https://doi.org/10.1016/S1076-6332(03)00723-2 Get rights and content

Abstract

Rationale and objectives

To develop and evaluate a two-stage computerized method that first detects suspicious regions on ultrasound images, and subsequently distinguishes among different lesion types.

Materials and methods

The first stage of detecting potential lesions was based on expected lesion shape and margin characteristics. After the detection stage, all candidate lesions were classified by a Bayesian neural net based on computer-extracted lesion features. Two separate tasks were performed and evaluated at the classification stage: the first classification task was the distinction between all actual lesions and false-positive detections; the second classification task was the distinction between actual cancer and all other detected lesion candidates (including false-positive detections). The neural nets were trained on a database of 400 cases (757 images), consisting of complex cysts and benign and malignant lesions, and tested on an independent database of 458 cases (1,740 images including 578 normal images).

Results

In the distinction between all actual lesions and false-positive detections, A_z values of 0.94 and 0.91 were obtained with the training and testing data sets, respectively. Sensitivity by patient of 90% at 0.45 false-positive detections per image was achieved for this detection-plus-classification scheme for the testing data set. Distinguishing cancer from all other detections (false-positives plus all benign lesions) proved to be more challenging, and A_z values of 0.87 and 0.81 were obtained during training and testing, respectively. Sensitivity by patient of 100% at 0.43 false-positive malignancies per image was achieved in the detection and classification of cancerous lesions for the testing dataset.

Conclusion

The results show promising performance of the computerized lesion detection and classification method, and indicate the potential of such a system for clinical breast ultrasound.

Section snippets

Materials

Two databases of sonographic images were used in this research. Patients were referred to diagnostic US for various reasons, including young age, dense breast parenchyma with microcalcifications, need for US-guided core biopsy, and for the distinction between cystic and solid lesions in abnormalities detected by mammography. The initial database was obtained from Northwestern University (Chicago, IL), and the second database was collected at the University of Chicago (Chicago, IL), both under

Results

In Figure 1, example images are shown with the computer output, detection points and candidate lesion outlines, together with the outlines by the radiologist for comparison. In the first two examples, the computer detected actual lesions correctly, while in the third example the two lesion candidates identified by the computer are false-positive detections. The comparison of computer outlines with human outlines was the subject of earlier studies (8). Indirectly, however, computer segmentation

Discussion

In this work, we presented a computerized method for detection and classification of lesions on breast US. Evaluation was performed for two schemes to assess whether actual lesions were adequately detected by the computer, and whether our classifier had sufficient potential to distinguish cancers from all other detected lesion candidates. The ability to correctly identify all of the few cancer cases within the testing database was promising, considering the difficulty of the task and

Acknowledgements

The authors thank Dr. Luz Venta and Michael Stern for their help in collecting the sonographic databases.

References (17)

D.R Chen et al.
Diagnosis of breast tumors with sonographic texture analysis using wavelet transform and neural networks
Ultrasound Med Biol
(2002)
C.E Metz
Basic principles of ROC analysis
Semin Nucl Med
(1978)
L.W Bassett et al.
Usefulness of mammography and sonography in women < 35 yrs old
Radiography
(1991)
T.M Kolb et al.
Occult cancer in women with dense breastsdetection with screening US-diagnostic yield and tumor characteristics
Radiology
(1998)
C.M Chen et al.
Breast lesions on sonogramscomputer-aided diagnosis with nearly setting-independent features and artificial neural networks
Radiology
(2003)
K Horsch et al.
Computerized diagnosis of breast lesions on ultrasound
Med Phys
(2002)
M.A Kupinski et al.
Computerized detection of mass lesions in digital mammography using radial gradient index filtering
Radiology
(1999)
K Drukker et al.
Computerized lesion detection on breast ultrasound
Med Phys
(2002)

There are more references available in the full text version of this article.

Cited by (104)

Breast ultrasound tumor image classification using image decomposition and fusion based on adaptive multi-model spatial feature fusion
2021, Computer Methods and Programs in Biomedicine
Breast cancer is a fatal threat to the health of women. Ultrasonography is a common method for the detection of breast cancer. Computer-aided diagnosis of breast ultrasound images can help doctors in diagnosing benign and malignant lesions. In this paper, by combining image decomposition and fusion techniques with adaptive spatial feature fusion technology, a reliable classification method for breast ultrasound images of tumors is proposed.
First, fuzzy enhancement and bilateral filtering algorithms are used to process the original breast ultrasound image. Then, various decomposition images representing the clinical characteristics of breast tumors are obtained using the original and mask images. By considering the diversity of the benign and malignant characteristic information represented by each decomposition image, the decomposition images are fused through the RGB channel, and three types of fusion images are generated. Then, from a series of candidate deep learning models, transfer learning is used to select the best model as the base model to extract deep learning features. Finally, while training the classification network, adaptive spatial feature fusion technology is used to train the weight network to complete deep learning feature fusion and classification.
In this study, 1328 breast ultrasound images were collected for training and testing. The experimental results show that the values of accuracy, precision, specificity, sensitivity/recall, F1 score, and area under the curve of the proposed method were 0.9548, 0.9811, 0.9833, 0.9392, 0.9571, and 0.9883, respectively.
Our research can automate breast cancer detection and has strong clinical utility. When compared to previous methods, our proposed method is expected to be more effective while assisting doctors in diagnosing breast ultrasound images.
An Integration of blockchain and AI for secure data sharing and detection of CT images for the hospitals
2021, Computerized Medical Imaging and Graphics
Deep learning, for image data processing, has been widely used to solve a variety of problems related to medical practices. However, researchers are constantly struggling to introduce ever efficient classification models. Recent studies show that deep learning can perform better and generalize well when trained using a large amount of data. Organizations such as hospitals, testing labs, research centers, etc. can share their data and collaboratively build a better learning model. Every organization wants to retain the privacy of their data, while on the other hand, these organizations want accurate and efficient learning models for various applications. The concern for privacy in medical data limits the sharing of data among multiple organizations due to some ethical and legal issues. To retain privacy and enable data sharing, we present a unique method that combines locally learned deep learning models over the blockchain to improve the prediction of lung cancer in health-care systems by filling the defined gap. There are several challenges involved in sharing that data while maintaining privacy. In this paper, we identify and address such challenges. The contribution of our work is four-fold: (i) We propose a method to secure medical data by only sharing the weights of the trained deep learning model via smart contract. (ii) To deal with different sized computed tomography (CT) images from various sources, we adopted the Bat algorithm and data augmentation to reduce the noise and overfitting for the global learning model. (iii) We distribute the local deep learning model wights to the blockchain decentralized network to train a global model. iv) We propose a recurrent convolutional neural network (RCNN) to estimate the region of interest (ROI) in theCT images. An extensive empirical study has been conducted to verify the significance of our proposed method for better prediction of cancer in the early stage. Experimental results of the proposed model can show that our proposed technique can detect the lung cancer nodules and also achieve better performance.
Deep learning radiomics in breast cancer with different modalities: Overview and future
2020, Expert Systems with Applications
Citation Excerpt :
With high sensitivity to calcification, mammographic examination is far better at detecting microcalcifications and clusters of calcifications, which are very important characterizations of breast cancer (Horsch et al., 2006). Ultrasound which uses non-ionizing radiation has been proven to be an important adjunct scanning tool to mammography for breast cancer (Corsetti et al., 2011; Drukker et al., 2004). Ultrasound is more sensitive in dense breasts and effective in distinguishing the cysts from solid masses, which are limited in mammography (Kelly et al., 2010).
Recent improvements in deep learning radiomics (DLR) extracting high-level features form medical imaging could promote the performance of computer aided diagnosis (CAD) for cancer. Breast cancer is the most frequent cancer among women and prospective achievements have been reported by CAD systems based on deep learning methods for breast imaging. In this paper, we aim to provide a comprehensive overview of the recent research efforts on DLR in breast cancer with different modalities and propose the future directions in this field. First, we respectively summarize and analyze the dataset, architecture, application and evaluation on DLR for breast cancer with three main imaging modalities, i.e., ultrasound, mammography, magnetic resonance imaging. Especially, we provide a survey on deep learning architectures exploited in breast cancer, including discriminative architectures and generative architectures. Then, we propose some potential challenges along with future research directions as references to the clinical treatment management and decision making utilizing such breast cancer CAD systems.
Breast ultrasound region of interest detection and lesion localisation
2020, Artificial Intelligence in Medicine
Citation Excerpt :
Fig. 1 shows BUS images with manual pre-selected ROIs marked with ‘+’ and ‘x’. Previous work attempted to automate the process of ROIs selection [10–13]. These methods were based on multi-stage image processing and/or machine learning approaches.
In current breast ultrasound computer aided diagnosis systems, the radiologist preselects a region of interest (ROI) as an input for computerised breast ultrasound image analysis. This task is time consuming and there is inconsistency among human experts. Researchers attempting to automate the process of obtaining the ROIs have been relying on image processing and conventional machine learning methods. We propose the use of a deep learning method for breast ultrasound ROI detection and lesion localisation. We use the most accurate object detection deep learning framework – Faster-RCNN with Inception-ResNet-v2 – as our deep learning network. Due to the lack of datasets, we use transfer learning and propose a new 3-channel artificial RGB method to improve the overall performance. We evaluate and compare the performance of our proposed methods on two datasets (namely, Dataset A and Dataset B), i.e. within individual datasets and composite dataset. We report the lesion detection results with two types of analysis: (1) detected point (centre of the segmented region or the detected bounding box) and (2) Intersection over Union (IoU). Our results demonstrate that the proposed methods achieved comparable results on detected point but with notable improvement on IoU. In addition, our proposed 3-channel artificial RGB method improves the recall of Dataset A. Finally, we outline some future directions for the research.
Artificial intelligence and interpretations in breast cancer imaging
2020, Artificial Intelligence in Medicine: Technical Basis and Clinical Applications
Medical decision-making in breast cancer image interpretation continues to evolve with synergistic advances in image acquisition systems (e.g., breast tomosynthesis), imaging protocols [e.g., multiparametric magnetic resonance imaging (MRI)], interpretation aids [artificial intelligence (AI)], and clinical tasks (use in response assessment). Advances in computers, in terms of both computing power and memory aspects, have led to a rapid increase in assessing the potential use of AI in various tasks in breast imaging, going beyond the initial use in computer-aided detection to include diagnosis, prognosis, response to therapy, and risk assessment, as well as in cancer discovery. AI methods are being developed for computer-aided detection (CADe) and diagnosis (CADx), for triaging (CADt), and for ultimate use as an autonomous reader, often with limited consideration for effect on radiologists' perception/cognitive performance and workflow. While the prospects of AI in breast cancer image interpretation are abundant and promising, they bring along challenges and trepidations. This chapter focuses on the role of AI in breast cancer image interpretation.
Automatic breast ultrasound image segmentation: A survey
2018, Pattern Recognition
Citation Excerpt :
Scientists do not definitely know what causes breast cancer yet, and only know some risk factors that can increase the likelihood of developing breast cancer: getting older, genetics, radiation exposure, dense breast tissue, alcohol consumption, etc. The key of reducing the mortality is to find signs and symptoms of breast cancer at its early stage by clinic examination [3]. Breast ultrasound (BUS) imaging has become one of the most important and effective modality for the early detection of breast cancer because of its noninvasive, nonradioactive and cost-effective nature [4]; and it is most suitable for large-scale breast cancer screening and diagnosis in low-resource countries and regions.
Breast cancer is one of the leading causes of cancer death among women worldwide. In clinical routine, automatic breast ultrasound (BUS) image segmentation is very challenging and essential for cancer diagnosis and treatment planning. Many BUS segmentation approaches have been studied in the last two decades, and have been proved to be effective on private datasets. Currently, the advancement of BUS image segmentation seems to meet its bottleneck. The improvement of the performance is increasingly challenging, and only few new approaches were published in the last several years. It is the time to look at the field by reviewing previous approaches comprehensively and to investigate the future directions. In this paper, we study the basic ideas, theories, pros and cons of the approaches, group them into categories, and extensively review each category in depth by discussing the principles, application issues, and advantages/disadvantages.

View all citing articles on Scopus

^☆: This work was supported in parts by USPHS grant CA89452, and U.S. Army Medical Research and Materiel Command 97-2445.

¹: Drs. Giger and Vyborny are shareholders in R2 Technology, Inc. (Los Altos, CA). It is the University of Chicago Conflict of Interest Policy that investigators disclose publicly actual or potential significant financial interest that would reasonably appear to be directly and significantly affected by the research activities.

View full text

Original investigationComputerized detection and classification of cancer on breast ultrasound1☆

Abstract

Rationale and objectives

Materials and methods

Results

Conclusion

Section snippets

Materials

Results

Discussion

Acknowledgements

Ultrasound Med Biol

Semin Nucl Med

Usefulness of mammography and sonography in women < 35 yrs old

Radiography

Occult cancer in women with dense breastsdetection with screening US-diagnostic yield and tumor characteristics

Radiology

Breast lesions on sonogramscomputer-aided diagnosis with nearly setting-independent features and artificial neural networks

Radiology

Computerized diagnosis of breast lesions on ultrasound

Med Phys

Computerized detection of mass lesions in digital mammography using radial gradient index filtering

Radiology

Computerized lesion detection on breast ultrasound

Med Phys

Original investigation
Computerized detection and classification of cancer on breast ultrasound¹☆