China
Africa
Explore chapters and articles related to this topic
Mammographic Screening and Breast Cancer Management – Part 2
Published in Sandeep Reddy, Artificial Intelligence, 2020
The question being asked by the clinician who orders a screening mammogram is relatively straight forward: Is there cancer? While other processes may be present, such as a breast cyst, the point of doing the screening mammogram is to answer “MAYBE” or “NO” to the question of cancer presence on the images. Indeed, every screening mammogram dictation finishes with a very brief summary that is a BI-RADS code. For a screening exam, the final report summary is either a BI-RADS 0,1 or 2. A “MAYBE” answer results in a BI-RADS 0 summary score which means the evaluation of the breast is incomplete as there may be a malignancy present. This score leads to the patient being called back for additional diagnostic mammogram views and/or targeted ultrasound to determine if a biopsy is indicated. A BI-RADS 1 (negative findings) or 2 (benign findings) summary score means the patient needs no further workup as no areas suspicious for cancer are detected. If the design of the algorithm can be focused and limited to detecting the morphologic changes that it has been taught are associated with malignancy, then it can potentially satisfy the clinical mission of the screening mammogram study.
Disease-Inspired Feature Design for Computer-Aided Diagnosis of Breast Cancer Digital Pathology Images
Published in de Azevedo-Marques Paulo Mazzoncini, Mencattini Arianna, Salmeri Marcello, Rangayyan Rangaraj M., Medical Image Analysis and Informatics: Computer-Aided Diagnosis and Therapy, 2018
Diagnosis of IDC occurs in two stages. First, the patient is sent for a mammogram, and the reviewing radiologist uses the Breast Imaging Reporting and Data System (BI-RADS) to note the existence and suspiciousness of any lesions. Usually, if a score of 4 or higher is reported for any of the lesions, the patient is sent for a biopsy to sample the tissue for further analysis. The sampled tissues are then examined under magnification by an anatomic pathologist. To quantify sample characteristics, grading and scoring systems are used by the pathologist to determine whether the tissue is diseased (diagnosis), what subclassifications of the disease are applicable [3], the level of tumor aggressiveness, as well as possible treatment options. The majority of these data are not discernable from imaging methods like mammograms alone. Therefore, pathology is the most informative modality for breast cancer diagnosis and provides critical information for patient treatment decisions.
The ratings paradigm
Published in Dev P. Chakraborty, Observer Performance Methods for Diagnostic Imaging, 2017
It is desirable that the rating scale be relevant to the radiologists' daily practice. This assures greater consistency—the fitting algorithms assume that the thresholds are held constant for the duration of the ROC study. Depending on the clinical task, a natural rating scale may already exist. For example, in 1992 the American College of Radiology developed the Breast Imaging Reporting and Data System (BI-RADS) to standardize mammography reporting.36 There are six assessment categories: category 0 indicates need for additional imaging; category 1 is a negative (clearly non-diseased) interpretation; category 2 is a benign finding; category 3 is probably benign, with short-interval follow-up suggested; category 4 is a suspicious abnormality for which biopsy should be considered; category 5 is highly suggestive of malignancy and appropriate action should be taken. The 4th edition of the BI-RADS manual37 divides category 4 into three subcategories 4A, 4B, and 4C and adds category 6 for a proven malignancy. The 3-category may be further subdivided into probably benign with a recommendation for normal or short-term follow-up and a 3+ category, and probably benign with a recommendation for immediate follow-up. Apart from categories 0 and 2, the categories form an ordered set with higher categories representing greater confidence in presence of cancer. How to handle the 0s and the 2s is the subject of some controversy, described next.
Convolutional Neural Networks Improve Radiologists’ Performance in Breast Cancer Screening for Vietnamese patients
Published in Applied Artificial Intelligence, 2022
Bui My Hanh, Le Tuan Linh, Nguyen Ngoc Cuong, Thanh Binh Nguyen, Luu Tien Doan, Chung Duy Le, Vu Tat Giao, Thi Ly Ly Ngo, Thi Hong Xuyen Hoang, Nguyen Duc Thang, Nguyen Tu Anh, Nguyen Duc Dan, Nguyen Viet Dung, Tran Vinh Duc, Quang H. Nguyen, Anh Nguyen, Nguyen Hoang Phuong
In Vietnam, the radiologists in many hospitals applied a classification system BI-RADS in 2013 (Breast Imaging Reporting and Data System) to classify a malignant level of abnormalities in breast cancer X-Ray images (Tan et al. 2004): BI-RADS 0: Incomplete assessment. Additional testing is needed.BI-RADS 1: The breast is normal, and no hurts were detected.BI-RADS 2: The breast has an abnormality but benign mass.BI-RADS 3: The breast has hurt, but the mass is most likely benign; the malignancy rate is less than 2%.BI-RADS 4: The breast has hurt with a high malignancy rate, i.e., from 2% to 95%.BI-RADS 5: The breast has hurt with a malignancy rate of over 95%.
A dual-mode deep transfer learning (D2TL) system for breast cancer detection using contrast enhanced digital mammograms
Published in IISE Transactions on Healthcare Systems Engineering, 2019
Kun Wang, Bhavika K. Patel, Lujia Wang, Teresa Wu, Bin Zheng, Jing Li
Finally, we discuss implications of our results for the current clinical practice on breast cancer. The Breast Imaging Reporting and Data System (BI-RADS®) is a standardized tool established by the American College of Radiology to guide radiologists to classify breast imaging findings. Cases classified into BI-RADS® Category 4 and 5 are those recommended for biopsy or surgical consultation. Currently, it is reported that only 25% of these cases result in a tissue diagnosis of malignant tumor within one year (Rosenberg et al., 2006). This number is too low. Our study demonstrated the promise of increasing this number to 80% by leveraging the proposed D2TL system to assist with radiologists’ decisions. This would positively impact the current patient care continuum.
A CNN based method for automatic mass detection and classification in mammograms
Published in Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, 2019
Ayelet Akselrod-Ballin, Leonid Karlinsky, Sharon Alpert, Sharbell Hashoul, Rami Ben-Ari, Ella Barkan
The objective of this paper is to introduce a novel algorithm for detection and classification of masses based on a powerful region-based convolutional networks approach. Classification is performed according to the breast imaging-reporting and data system (BI-RADS) score (Narvaez et al. 2010). The BI-RADS score ranges from 0 to 6 and is defined as (0 more information is needed,1 negative, 2 benign finding, 3 probably benign <2% likelihood of cancer, 4 suspicious abnormality, 5 highly suggestive of malignancy and 6 proven malignancy). This study dealt only with scores 1 to 5 based on radiological features and without correlation with clinical information. We demonstrate our results on tumour detection and classification where due to the small amount of training data in some of the score classes, we include the three major clinical classes of normal{1}, benign {2} and lesions that need further workup {3,4,5}. Figure 1, demonstrates examples of masses with different BI-RADS score.