Background Mammographic density improves the accuracy of breast cancer risk models. However, the use of breast density is limited by subjective assessment, variation across radiologists, and restricted data. A mammography-based deep learning (DL) model may provide more accurate risk prediction. Purpose To develop a mammography-based DL breast cancer risk model that is more accurate than established clinical breast cancer risk models. Materials and Methods This retrospective study included 88 994 consecutive screening mammograms in 39 571 women between January 1, 2009, and December 31, 2012. For each patient, all examinations were assigned to either training, validation, or test sets, resulting in 71 689, 8554, and 8751 examinations, respectively. Cancer outcomes were obtained through linkage to a regional tumor registry. By using risk factor information from patient questionnaires and electronic medical records review, three models were developed to assess breast cancer risk within 5 years: a risk-factor-based logistic regression model (RF-LR) that used traditional risk factors, a DL model (image-only DL) that used mammograms alone, and a hybrid DL model that used both traditional risk factors and mammograms. Comparisons were made to an established breast cancer risk model that included breast density (Tyrer-Cuzick model, version 8 [TC]). Model performance was compared by using areas under the receiver operating characteristic curve (AUCs) with DeLong test (P < .05). Results The test set included 3937 women, aged 56.20 years ± 10.04. Hybrid DL and image-only DL showed AUCs of 0.70 (95% confidence interval [CI]: 0.66, 0.75) and 0.68 (95% CI: 0.64, 0.73), respectively. RF-LR and TC showed AUCs of 0.67 (95% CI: 0.62, 0.72) and 0.62 (95% CI: 0.57, 0.66), respectively. Hybrid DL showed a significantly higher AUC (0.70) than TC (0.62; P < .001) and RF-LR (0.67; P = .01). Conclusion Deep learning models that use full-field mammograms yield substantially improved risk discrimination compared with the Tyrer-Cuzick (version 8) model. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Sitek and Wolfe in this issue.

https://pubs.rsna.org/doi/pdf/10.1148/radiol.2019182716

A Deep Learning Mammography-based Model for Improved Breast Cancer Risk Prediction

Biotelemetry—the process of conveying data from a transmitter-attached animal to a data collection site—has received increasing awareness from fisheries researchers. Prior to biotelemetry data collection, it is imperative that researchers are aware of and understand the possible effects that transmitter presence and attachment procedures may have on ‘normal’ fish behavior and physiology. To allow successful transmitter attachment, numerous methods to anesthetize the study fish, with varied impact and effectiveness, may be employed. Following anesthetization, three standard methods of transmitter attachment have been developed—external attachment, intragastric insertion, and surgical implantation. Although each method has advantages and disadvantages, their success largely depends on factors such as the species, environment, fish and transmitter size, and duration of the telemetry study. Additionally, each method of attachment can affect experimental fish physiology and behavior in varying ways. After desc...

http://www3.carleton.ca/fecpl/pdfs/Bridger%20and%20Booth%20PDF.pdf

The Effects of Biotelemetry Transmitter Presence and Attachment Procedures on Fish Physiology and Behavior

Passive acoustic telemetry provides an important tool to study the spatial ecology and behav- iour of organisms in marine and freshwater systems, but understanding the detection range of acoustic receivers is critical for interpreting acoustic data and establishing receiver spacing to maximize study efficiency. This study presents a comprehensive review of how acoustic detection range has been considered and assessed to date, summarizes important variables to monitor when determining the detection range of a receiver array, and provides recommendations to account for detection range during experimental design, analysis and data interpretation. A total of 378 passive acoustic telemetry studies (1986-2012) were scored against a set of pre- defined criteria to provide a standardized assessment of how well detection range was accounted for, from a maximum possible score of 45. Scores ranged from 0 to 39 (11.1 ± 0.4; mean ± 1 SE). Over the past decade mean scores have been consistently between 6.7 and 12.9 which indicates that detection range has not been adequately considered in most contemporary acoustic telemetry studies. Given the highly variable nature of detection range over space and time, it is necessary to create a culture of detection range testing among the scientific community. For robust telemetry studies it is recommended that consideration of detection range should be given a greater focus within study design, execution and data analysis. To aid array design in new systems, short-term detection range tests should be conducted in the most representative area of the study system prior to deployment. As well, fixed distance

A review of detection range testing in aquatic passive acoustic telemetry studies

Early approaches to surgical implantation of electronic tags in fish were often through trial and error, however, in recent years there has been an interest in using scientific research to identify techniques and procedures that improve the outcome of surgical procedures and determine the effects of tagging on individuals. Here we summarize the trends in 108 peer-reviewed electronic tagging effect studies focused on intracoleomic implantation to determine opportunities for future research. To date, almost all of the studies have been conducted in freshwater, typically in laboratory environments, and have focused on biotelemetry devices. The majority of studies have focused on salmonids, cyprinids, ictalurids and centrarchids, with a regional bias towards North America, Europe and Australia. Most studies have focused on determining whether there is a negative effect of tagging relative to control fish, with proportionally fewer that have contrasted different aspects of the surgical procedure (e.g., methods of sterilization, incision location, wound closure material) that could advance the discipline. Many of these studies included routine endpoints such as mortality, growth, healing and tag retention, with fewer addressing sublethal measures such as swimming ability, predator avoidance, physiological costs, or fitness. Continued research is needed to further elevate the practice of electronic tag implantation in fish in order to ensure that the data generated are relevant to untagged conspecifics (i.e., no long-term behavioural or physiological consequences) and the surgical procedure does not impair the health and welfare status of the tagged fish. To that end, we advocate for (1) rigorous controlled manipulations based on statistical designs that have adequate power, account for inter-individual variation, and include controls and shams, (2) studies that transcend the laboratory and the field with more studies in marine waters, (3) incorporation of knowledge and techniques emerging from the medical and veterinary disciplines, (4) addressing all components of the surgical event, (5) comparative studies that evaluate the same surgical techniques on multiple species and in different environments, (6) consideration of how biotic factors (e.g., sex, age, size) influence tagging outcomes, and (7) studies that cover a range of endpoints over ecologically relevant time periods.

Advancing the surgical implantation of electronic tags in fish: a gap analysis and research agenda based on a review of trends in intracoelomic tagging effects studies

Purpose To develop a deep learning (DL) algorithm to assess mammographic breast density. Materials and Methods In this retrospective study, a deep convolutional neural network was trained to assess Breast Imaging Reporting and Data System (BI-RADS) breast density based on the original interpretation by an experienced radiologist of 41 479 digital screening mammograms obtained in 27 684 women from January 2009 to May 2011. The resulting algorithm was tested on a held-out test set of 8677 mammograms in 5741 women. In addition, five radiologists performed a reader study on 500 mammograms randomly selected from the test set. Finally, the algorithm was implemented in routine clinical practice, where eight radiologists reviewed 10 763 consecutive mammograms assessed with the model. Agreement on BI-RADS category for the DL model and for three sets of readings-(a) radiologists in the test set, (b) radiologists working in consensus in the reader study set, and (c) radiologists in the clinical implementation set-were estimated with linear-weighted κ statistics and were compared across 5000 bootstrap samples to assess significance. Results The DL model showed good agreement with radiologists in the test set (κ = 0.67; 95% confidence interval [CI]: 0.66, 0.68) and with radiologists in consensus in the reader study set (κ = 0.78; 95% CI: 0.73, 0.82). There was very good agreement (κ = 0.85; 95% CI: 0.84, 0.86) with radiologists in the clinical implementation set; for binary categorization of dense or nondense breasts, 10 149 of 10 763 (94%; 95% CI: 94%, 95%) DL assessments were accepted by the interpreting radiologist. Conclusion This DL model can be used to assess mammographic breast density at the level of an experienced mammographer. © RSNA, 2018 Online supplemental material is available for this article . See also the editorial by Chan and Helvie in this issue.

https://pubs.rsna.org/doi/pdf/10.1148/radiol.2018180694

Mammographic Breast Density Assessment Using Deep Learning: Clinical Implementation

Purpose To compare the classification of breast density with two automated methods, Volpara (version 1.5.0; Matakina Technology, Wellington, New Zealand) and Quantra (version 2.0; Hologic, Bedford, Mass), with clinical Breast Imaging Reporting and Data System (BI-RADS) density classifications and to examine associations of these measures with breast cancer risk. Materials and Methods In this study, 1911 patients with breast cancer and 4170 control subjects matched for age, race, examination date, and mammography machine were evaluated. Participants underwent mammography at Mayo Clinic or one of four sites within the San Francisco Mammography Registry between 2006 and 2012 and provided informed consent or a waiver for research, in compliance with HIPAA regulations and institutional review board approval. Digital mammograms were retrieved a mean of 2.1 years (range, 6 months to 6 years) before cancer diagnosis, with the corresponding clinical BI-RADS density classifications, and Volpara and Quantra density estimates were generated. Agreement was assessed with weighted κ statistics among control subjects. Breast cancer associations were evaluated with conditional logistic regression, adjusted for age and body mass index. Odds ratios, C statistics, and 95% confidence intervals (CIs) were estimated. Results Agreement between clinical BI-RADS density classifications and Volpara and Quantra BI-RADS estimates was moderate, with κ values of 0.57 (95% CI: 0.55, 0.59) and 0.46 (95% CI: 0.44, 0.47), respectively. Differences of up to 14% in dense tissue classification were found, with Volpara classifying 51% of women as having dense breasts, Quantra classifying 37%, and clinical BI-RADS assessment used to classify 43%. Clinical and automated measures showed similar breast cancer associations; odds ratios for extremely dense breasts versus scattered fibroglandular densities were 1.8 (95% CI: 1.5, 2.2), 1.9 (95% CI: 1.5, 2.5), and 2.3 (95% CI: 1.9, 2.8) for Volpara, Quantra, and BI-RADS classifications, respectively. Clinical BI-RADS assessment showed better discrimination of case status (C = 0.60; 95% CI: 0.58, 0.61) than did Volpara (C = 0.58; 95% CI: 0.56, 0.59) and Quantra (C = 0.56; 95% CI: 0.54, 0.58) BI-RADS classifications. Conclusion Automated and clinical assessments of breast density are similarly associated with breast cancer risk but differ up to 14% in the classification of women with dense breasts. This could have substantial effects on clinical practice patterns. (©) RSNA, 2015 Online supplemental material is available for this article.

Comparison of Clinical and Automated Breast Density Measurements: Implications for Risk Prediction and Supplemental Screening

Using scuba in August and September 1997 and 1998, we surgically implanted acoustic transmitters (Vemco V16 series) in 6 greenspotted rockfish Sebastes chlorostictus and 16 bocaccio S. paucispinis on the flank of Soquel Canyon in Monterey Bay, California. In 1997 we used longline gear to capture and tag greenspotted rockfish, and in 1998 we worked with a commercial fisherman who used modified trolling gear to help us capture and tag bocaccio. Each year, fish were captured at depths of 100–200 m and reeled up to a depth of approximately 20 m. This depth was chosen to reduce temperature and pressure stress caused by bringing fish to the surface. A team of scuba divers descended to the 20-m depth and anesthetized and surgically implanted acoustic transmitters in both species of rockfish. Tagged fish were released on the bottom at the location of catch. Several weeks after tagging, we used the Delta submersible to place an array of recording receivers on the seafloor, which enabled tracking of horizo...

John N. Heine

Papers

Comparison of Clinical and Automated Breast Density Measurements: Implications for Risk Prediction and Supplemental Screening

Techniques for Tagging and Tracking Deepwater Rockfishes