Susan G Komen  
I've Been Diagnosed With Breast Cancer Someone I Know Was Diagnosed Share Your Story Join Us And Stay Informed Donate To End Breast Cancer
    Home > Research & Grants > Grants Program > Research Grants > Research Grants Awarded > Abstract
    Awarded Grants
    The Cognitively-Based Hypermedia System for CAD-Assisted Mammography Interpretation

    Scientific Abstract:
    The Cognitively-Based Hypermedia System for CAD-Assisted Mammography Interpretation Background: Effective training of radiology residents and computer-assisted diagnosis (CAD) are two solutions to the early detection of breast cancer. The existing radiology training programs are time-consuming, cost-ineffective and lacking in feedbacks. Moreover, they are somewhat haphazard and inconsistent in curriculum, quality of presentation, quality and quantity of practice, and instructors’ time. Therefore, this research project intends to propose an effective, interactive, and standardized Web-based hypermedia tutoring system in CAD-assisted mammography interpretation based on findings in cognitive model and processes, and statistically significant evidences in empirical studies. Objective/Hypothesis: The objective of the project is to identify the problem-solving model and cognitive components underlying CAD-assisted mammography interpretation expertise. It also designs and creates a web-based hypermedia prototype tutoring system on the basis of cognitive science and empirical evidences. Three main hypotheses are as follows: (1) CAD-assisted mammography interpretation may greatly reduce the differences in performance measures, diagnostic accuracy, error types, problem-solving strategies, and control processes among novice, intermediate and expert radiologists. (2) The efficacy of the proposed cognitively-based intelligent tutoring system may be significantly better than the other existing systems. (3) The integration of CAD into training system may significantly enhance performance in mammography interpretation. Specific Aims: (1) To construct a problem-solving model of CAD-assisted mammography interpretation, (2) to analyze and identify if and how CAD greatly reduces the novice-expert differences in radiological observation, findings, diagnosis, diagnostic accuracy, reasoning strategies, error types, problem-solving operators and control processes, (3) to design an intelligent tutoring system for training both radiologist residents and medical students in CAD-assisted mammography interpretation, using the results of expert knowledge study, content and task analysis and (4) to develop a prototype Web-based training environment embedded in cognitive tools with hypermedia authoring software. Study Design: Two parallel studies will be conducted to compare the differences between non-CAD and CAD-assisted mammography interpretation across levels of expertise. Twenty cases digitized at 60-micron resolution will be selected from established database. Each case is comprised of a de-identified coded clinical history and a set of mammograms. Thinking protocols will be coded along three axes: knowledge states, problem-solving operators, and control processes. Also error analyses will be conducted for pattern recognition. ANOVAs and Chi-square analyses will be conducted to measure performance differences among groups in terms of radiological findings, observations, and diagnoses, as well as diagnostic accuracy, problem-solving strategies, frequency of control process use, frequency of problem-solving operator use, requests for additional medical information, and errors. The system will be designed and created based on protocol analysis and followed-up task and content analysis. The system consists of several modules with 1,000 cases. The link-rich domain knowledge module will cover the knowledge base in breast anatomy, physiology, pathology and projective geometry of radiography. The schema module will present the cognitive model in mammography interpretation. In the main module of simulated diagnosis, the computer tutor will scaffold the user through the overall process of reading clinical histories, positioning images, identifying and marking observations and findings, using CAD schemes as a second reader, and providing diagnosis and subsequent examinations. The CAD module will provide various CAD schemes. The library module will extend practice to a network access database of ill-defined cases. It allows rapid access and comparison of diagnosis of CAD reader and radiologists. Potential Outcomes and Benefits of the Research: The research may well find out the cognitive model and processes of radiologists. The results will be used to construct a system for training both residents and medical students in CAD-assisted mammography interpretation. It will help standardize resident training programs, making them more interactive, accessible and adaptable. The increasing accuracy in mammography interpretation will greatly benefit the early detection of breast cancer.

    Lay Abstract:
    The Cognitively-Based Hypermedia System for CAD-Assisted Mammography Interpretation The training of radiological residents and medical students in mammography interpretation requires more attentions due to the seriousness and scope of breast cancer, and the difficulties and complexity in interpreting mammograms. The existing residency training programs are time-consuming, cost-ineffective and lacking in feedbacks. Moreover, they are somewhat haphazard and inconsistent in curriculum, quality of presentation, quality and quantity of practice, and instructor’s time. The problems in radiology training programs suggest that more efficient, interactive, and standardized system should be established to improve radiological expertise in mammography interpretation. Therefore, this research project proposes an intelligent tutoring system based on cognitive science and empirical evidences. The objective of the project is to identify the problem-solving model and cognitive factors underlying expertise in mammography interpretation with computer-assisted diagnosis (CAD) as a second reader. It also designs and creates a computer-based environment on the basis of these results, allowing radiologists and medical students to acquire radiological expertise through intensive simulated diagnosis and extensive problem-solving practice. About 1,000 cases will be used in the system, including a network access database of ill-defined cases. Moreover, the system will integrate CAD schemes to help learn how to increase perception ability of human readers. The proposed system will be significantly better than the other existing training environments in its efficacy, interactivity and standardization. Firstly, the integration of CAD may greatly enhance performance in mammography interpretation and greatly reduce the differences in performance measures, diagnostic accuracy, error types, problem-solving strategies, and control processes among novice, intermediate and expert radiologists. Secondly, the proposed system will provide very interactive interface and modules, simulating tutor-user communication. The user can have immediate idea of what he is mistaken in his visual perception and problem solving, as well as the guide to how to solve the problem in a right way. Thirdly, the repeated trials in observation, findings and diagnosis give the user a lot of practice and experience. Their knowledge and skills will be largely improved in a short period of time. Fourthly, the Web-based environment will make it very convenient for the user to learn and practice. They will have little limitation in time and space. Fifthly, the large database of various types of ill-defined cases will make it possible for the user to compare their own diagnosis with that from CAD and radiologists. The multiple observations, comparisons and diagnosis on the same case will greatly improve the user’s perception and cognitive skills. Sixthly, the research may well find out the cognitive processes and knowledge base of radiologist professionals at different expertise levels. The results will be used to construct an intelligent system for training both residents and medical students for CAD-assisted mammography interpretation. It will help standardize resident training. Finally, the increasing accuracy in mammography interpretation through using this proposed intelligent training system will greatly benefit the early detection of breast cancer.