MARCH 2012 AJR INBRIEF
Welcome to the March 2012 AJR InBrief. This issue features a compelling case for implementing CAD by Dr. Robert Nishikawa, a short list noteworthy articles from the March issue, a round-up of AJR articles making viral buzz around the web, and the latest CME and SAM offerings.
THE MARCH '12 AJR SHORT LIST
The March 2012 AJR features a timely and relevant selection of articles from the highlighted section on CARDIOPULMONARY IMAGING. These articles include original research on “Incidental Myocardial Infarct on Conventional Nongated CT: A Review of the Spectrum of Findings With Gated CT and Cardiac MRI Correlation.” and “Coronary Artery Calcium Can Predict All-Cause Mortality and Cardiovascular Events on Low-Dose CT Screening for Lung Cancer.” This issue also showcases the following noteworthy articles:
Click here for the full March 2012 AJR Table of Contents.
BEHIND THE JOURNAL WITH...DR. ROBERT NISHIKAWA
What can radiologists do to more effectively detect potential cancers? Does it take more training? More experience? In the March 2012 AJR , Dr. Robert Nishikawa and his colleagues from the University of Chicago, make the case for using CAD as a powerful tool for identifying lesions and refining image quality. In this edition of AJR InBrief , Dr. Nishikawa shares his thoughts on the benefits of CAD, its limitations, and his recommendations for adding CAD to initial imaging protocols.
How does CAD help increase the sensitivity for screening mammography?
All radiologists miss lesions, and mammography is no exception. Breast cancer can be missed on a mammogram because either the radiologist fails to identify the lesion because of an incomplete search of the images; or they spot the lesion but they fail to recognize it as a cancer and dismiss it. CAD will help with the former. A computer will always perform a complete, detailed search. In most cases, if radiographic signs of a cancer are present, the computer will mark the area as suspicious. When the radiologists reviews the CAD findings, if he or she can recognize that the computer has marked a cancer, then the patient can correctly be recalled.
How does a radiologist’s experience and exposure to mammography affect his or her ability to use CAD effectively?
I believe that those [radiologists] with more experience will derive the most benefit (i.e., an increase in sensitivity without a large increase in recall rate) because those radiologists will be better able to distinguish computer-false detections from computer-true detections. Less experienced radiologists may have difficulty in rejecting computer-false detections.
Is there a risk of radiologists becoming “lazy” by relying on CAD instead of refining their reading skills with experience and repetition?
At the current level of performance, CAD has a reasonably high sensitivity, but a relatively high false detection rate. So a radiologist could not completely rely on CAD. It is possible that the radiologist relies on the computer to do the search and then just reviews the CAD prompts. The data for this type of reading indicates that it is suboptimal compared to using CAD as a “second reader,” where the radiologist makes an initially reading and then reviews the CAD prompts. This paradigm is the way that CAD should be used, as that is the only method of reading approved by the FDA. The non-approved method may reduce the time a radiologist spends interpreting a case, but the radiologist will miss more cancers. So, there is a risk that a radiologist may read in the non-approved method. More training for radiologists on how to use CAD properly and the dangers of off-label use may reduce this risk.
What are some of the limitations for using CAD?
There are two major limitations to using CAD. The first is an increase in recall rate. The second is an increase in reading time. The absolute increase in recall rate is small, but as a percentage of the non-CAD recall rate, it is comparable to the percentage increase in sensitivity. The net result is that the positive predictive value is virtually unchanged, with a 10% increase in sensitivity when CAD is used. Since the radiologist has to effectively read the image twice, once without and then once with CAD, the reading will be longer.
There is an alternative reading paradigm being developed by Nico Karssemeijer, where both these limitations may be lessened or eliminated. In his studies, radiologists only view CAD prompts in specific areas where they finds a suspicious lesion. The radiologist queries the computer for any computer prompts at a specific location. If there is a prompt present, then the radiologist can be more confident that a malignancy is present. If there is no prompt, then the radiologist can more confidently decide not to recall the patient to work up that area. Nico has shown that the radiologists’ improvement in performance is higher using his method than the “second reader” method.
Do you think CAD will increase the level of false-positive readings?
CAD will definitely increase the level of false-positive readings. Not only do radiologists miss cancers, but they also miss benign lesions and pseudolesions (i.e., those caused by superposition of normal tissue). CAD will mark all these types of lesions. If CAD marks a benign or pseudolesion that the radiologist initially overlooked, then when reviewing the CAD marks, the radiologist may recall the woman for a benign or pseudolesion identified by CAD. The absolute increase in recall rate is small, but as a percentage of the non-CAD recall rate it is comparable to the percentage increase in sensitivity.
Do you believe that CAD should be added to the initial protocol for mammograms?
From a patient-centered outcome point of view, CAD is beneficial and should be used by radiologists. However, I do believe that radiologists need to have better training to use CAD more effectively. The major conclusion of our study was that radiologists ignore 70% of true-positive CAD prompts.
Ultimately, do you believe that CAD has the ability to impact the overall health and mortality for patients and the future developments in the field of mammography? How?
Used with mammography, CAD can lead to earlier detection of breast cancer by reducing radiologists’ miss rate. This potentially can lead to a reduction in morbidity and mortality. No studies have been conducted to examine this issue. I believe that morbidity studies can and will be done in the future, but a clinical mortality study is very unlikely. Modeling may be the best approach to examine whether CAD can reduce mortality.
The trend in breast imaging is to [use] 3D [imaging] through either CT or tomosynthesis. While both of these modalities can provide more diagnostic information, they produce many images that radiologists need to review. Comparing left to right breasts, different projections of the same breast, and images taken on different days is much more problematic with a 3D image set than a 2D image set. Further, searching for clustered microcalcifications in a breast volume is more difficult than in a 2D image. The types of problems associated with large datasets are potentially solved or reduced in difficulty through the use of CAD or CAD-driven image processing and display techniques.
THE SOCIAL MEDIA BUZZ
The AJR is being “liked,” “tweeted,” and shared from a variety of social media outlets. Here’s a sampling of the AJR articles with the highest viral buzz this past month:
CMEs AND SAMs The AJR is being “liked,” “tweeted,” and shared from a variety of social media outlets. Here’s a sampling of the AJR articles with the highest viral buzz this past month:
- Chemoembolization Practice Patterns and Technical Methods Among Interventional Radiologists: Results of an Online Survey
- False-Positive Lesions Mimicking Breast Cancer on FDG PET and PET/CT
- MDCT Arthrography of the Shoulder With Datasets of Isotropic Resolution: Indications, Technique, and Applications
- Imaging of Limb Salvage Surgery
- CT of Coronary Heart Disease: Part 1, CT of Myocardial Infarction, Ischemia, and Viability*
- CT of Coronary Heart Disease: Part 2, Dual-Phase MDCT Evaluates Late Symptom Recurrence in ST-Segment Elevation Myocardial Infarction Patients After Revascularization*
- Chest Radiography in the ICU: Part 1, Evaluation of Airway, Enteric, and Pleural Tubes*
- Chest Radiography in the ICU: Part 2, Evaluation of Cardiovascular Lines and Other Devices*
*These articles are also available for SAM credit.
TAKE OUR SURVEY
Love AJR ? Want to make a few suggestions for making the journal even better? You're in luck. You can share your thoughts and opinions with our Winter 2012 Mini-Survey. Click here to take our survey and submit your feedback. The results will be revealed in the May 2012 edition of AJR InBrief.
SOCIAL MEDIA
Stay abreast of ARRS news and updates by selecting the "Like" button on Facebook at http://www.facebook.com/ AmericanRoentgenRaySociety and following us on Twitter at @ARRS_Radiology.