Radiologists' performance in digital mammography and breast tomosynthesis

描述(申请人提供)：数字乳腺断层合成的临床可行性：观察者性能分析背景：数字乳腺断层合成(DBT)被认为是乳腺癌成像的最新和最有前途的发展。在临床初步研究中，DBT表现出比投影乳房X光摄影更高的敏感性和特异性，后者是目前乳腺癌筛查的首选工具。重要的是，与现有的数字乳房X光摄影(DM)系统相比，DBT的溢价相对较低。由于这些原因，人们普遍认为，在乳腺癌的诊断中，DBT将取代DM；一些人进一步认为，在乳腺癌的筛查中，DBT可能会取代乳房X线摄影。实施DBT的一个实际问题是放射科医生的审查时间，因为DBT产生的图像大约是乳房X光检查的50倍。审查时间的增加可能会显著影响乳腺癌筛查能力，从而影响筛查的机会。目的：我们的目的是通过比较放射科医生在通过时间控制观察实验解释DBT图像和DM图像以检测乳腺癌方面的表现，来洞察DBT作为人群乳腺癌筛查方法的可能成功。时间控制的观看条件将提供无偏见的感知观看情况，以比较与2D DM相比，3D DBT是否有助于对乳腺癌的感知识别。我们还将记录放射科医生查看DBT和DM病例时的眼睛位置。眼睛位置分析将提供一个严格的衡量标准，使放射科医生的表现能够进行定量测量。这些将有助于了解审查问题的现状，以便在今后的研究中更好地处理这些问题。特定目标(SA)：SA1：开发一个在时间控制观看条件下显示DBT和DM病例的平台。SA2：比较DBT和DM在时间控制观察条件下放射科医生对乳腺癌的检测性能，并分析眼位。研究设计：SA1：任务1：我们将修改现有的显示硬件和软件，以允许同样短的显示时间(2 S和4 S)用于DBT和DM图像。任务2：我们将为放射科医生设计一个图形用户界面，以便在每次时间控制的观看后报告他们的决定。SA2：任务1：经验丰富的MQSA认证放射科医生将选择一组配对的100个乳房的DBT和DM图像。任务2：四名经验丰富的MQSA认证放射科医生将在短时间控制的演示期间审查图像，同时跟踪他们的眼睛位置。任务3：我们将对眼位数据进行处理。任务4：我们将比较DBT和DM的放射科医生对乳腺癌的检测性能和对眼位数据的分析。公共卫生相关性：该提案旨在调查数字乳房断层合成(DBT)作为人群筛查方式的潜力。我们将重点比较放射科医生在堆叠模式下快速解释DBT横断面图像和DM投影图像的性能，以使用短时间控制观察实验检测乳腺癌。

DESCRIPTION (provided by applicant): Clinical Feasibility of Digital Breast Tomosynthesis: Analysis of Observer Performance Background: Digital breast tomosynthesis (DBT) is considered the latest and most promising development in breast cancer imaging. In clinical pilot studies DBT has demonstrated higher sensitivity and specificity over projection mammography, which is presently the breast cancer screening tool of choice. Importantly, DBT will be provided at a relatively modest premium compared to existing digital mammography (DM) systems. For these reasons it has been widely suggested that DBT will replace DM in the diagnosis of breast cancer; some further believe that DBT may replace mammography in screening breast cancer. A practical concern for implementation of DBT is the radiologists' review time because DBT generates approximately 50 times more images than mammography. An increase in review time might significantly impact the breast cancer screening capacity and thus access to screening. Objective: Our aim is to obtain insight into the likely success of DBT as the population breast cancer screening modality of choice by comparing the radiologists' performance to interpret DBT images with DM images for the detection of breast cancer using time-controlled viewing experiments. The time-controlled viewing conditions will provide an unbiased perceptual viewing situation to compare whether 3D DBT facilitates perceptual recognition of breast cancers compared to 2D DM. We will also record the eye positions of radiologists as they view DBT and DM cases. The eye position analyses will provide a rigorous metric that will enable the quantitative measurement of the radiologists' performance. These will help to understand the current status of review issues so that they can be better addressed in future studies. Specific Aims (SA): SA1: Develop a platform to display DBT and DM cases under time-controlled viewing conditions. SA2: Compare for DBT and DM the radiologists' breast cancers detection performance under time-controlled viewing conditions and analyze the eye positions. Study Design: SA1: Task 1: We will modify our existing display hardware and software to allow equally short presentation times (2 s and 4 s) for DBT and DM images. Task 2: We will design a graphical user interface for the radiologists to report their decisions after each time-controlled viewing. SA2: Task 1: An experienced MQSA-certified radiologist will select a set of paired DBT and DM images of 100 breasts. Task 2: Four experienced MQSA-certified radiologists will review the images during short time-controlled presentations while their eye positions are tracked. Task 3: We will process the eye-position data. Task 4: We will compare the radiologists' breast cancer detection performance and the analyses on the eye-position data between DBT and DM. PUBLIC HEALTH RELEVANCE: This proposal aims to investigate the potential of digital breast tomosynthesis (DBT) as a population screening modality. We will focus on comparing radiologists' performance to rapidly interpret DBT cross-sectional images in stack mode with DM projection images for the detection of breast cancer using short time-controlled viewing experiments.