Breast cancer is the most common cancer among women worldwide, and the second leading cause of cancer-related deaths among this population in the United States. As with many other types of cancer, early detection can have a significant impact in patient prognosis, and the recommendation of mammography screening to the general women population older than 40 years of age has saved thousands of lives. However, the reading of a mammogram is a very complex perceptual and cognitive task, and the applicant has previously shown that, under the current training guidelines, at the end of their residency, novice radiologists are no better at detecting breast cancer than breast technologists, who do not receive any formal training in how to detect this disease. Moreover, a fraction of these novice radiologists will become generalists, namely, they will be board certified to read a lot of different images, including mammograms. The lack of specialized training in breast imaging is reflected in the performance of these general radiologists, who not only detect fewer breast cancers but also detect fewer early-staged breast lesions when compared to better trained breast radiologists. This problem is acute amongst African-American women, who, in general, due to socio-economical status, receive their mammograms at community-based centers or at outreach clinics, facilities that tend to be staffed by general radiologists. In this project we propose to build upon the framework of a previously designed and evaluated computer-based tutoring system, named SlideTutor, which has been developed for another visual domain within Medicine, namely, Pathology. In that domain, use of SlideTutor has been shown to produce significant improvements in diagnostic reasoning among novice pathologists, and retention of the learned material has been observed over time. Thus, it is our goal to use a similar framework to develop a computer-based cognitive tutoring system that can teach generalists and novice radiologists (i) how to detect early-staged breast cancer (thus yielding improvements in overall sensitivity);and (ii) how to reduce the number of unnecessary biopsy recommendations (therefore improving specificity). Hence, it is our hypothesis that use of cognitive tutoring will. If successful, such intervention should have a significant impact in health care delivery to socio-economically disadvantage populations, and to reduce health care disparities between minority populations (such as African-Americans) and more affluent populations. In addition, in order to reach as wide a number of radiologists as possible, our computer-based tutoring system will be placed on the Internet, where it will be available, free of charge, to any radiologists who register to receive training in this task. Currently, the missed rates at mammography screenings are between 10 and 30 percent. In this project, we will develop a computer based tutoring system to train radiologists how to detect early signs of breast cancer. Our system will be deployed over the internet and will be available free of charge.
Currently, the missed rates at mammography screenings are between 10 and 30 percent. In this project, we will develop a computer based tutoring system to train radiologists how to detect early signs of breast cancer. Our system will be deployed over the internet and will be available free of charge.
|Mello-Thoms, Claudia; Mello, Carlos A B; Medvedeva, Olga et al. (2012) Perceptual analysis of the reading of dermatopathology virtual slides by pathology residents. Arch Pathol Lab Med 136:551-62|
|Chakraborty, Dev P; Yoon, Hong-Jun; Mello-Thoms, Claudia (2012) Application of threshold-bias independent analysis to eye-tracking and FROC data. Acad Radiol 19:1474-83|