Considerable space was given to potential developments in computer-aided detection (CAD) in a variety of breast imaging techniques at the computer exhibits area of the infoRAD demonstration hall at this week's Radiological Society of North America meeting in Chicago.
Perhaps fitting because it's their home town, the Department of Radiology, Committee on Medical Physics, at The University of Chicago have been very busy in presenting their work in an assortment of uses of CAD ("e" and "x") for breast imaging.
The first of the exhibits explores "An Advanced Workstation for Implementing Computer-Aided Diagnosis (CADx) in Diagnostic Breast Imaging." The exhibit notes that while conventional CAD systems (or CADe - computer-aided detection) are commonly available, commercially the CADx have yet to be put to use in clinical settings. For their purposes, an assortment of researchers at U. of Chicago developed two different types of workstations which represent two different approaches of "conveying the output of the CADx scheme to assist radiologists in making biopsy recommendations," according to a summary of the work.
The first of the workstations is multi-modality which has the ability to evaluate mammograms, sonograms, and dynamic breast MRI images. The second workstation looks at clustered calcifications and is geared for mammography. It demonstrates the effects of a "novel method for training the CADx scheme," according to the researchers.
Though the investigation of the workstations is ongoing, a number of new things are already being learned about the CADx systems - such as the development of a special interface for breast MRI, new methods for teaching an artificial neural network to increase the precision of the systems, and making use of prior mammograms and radiology reports from the clinical population to make assessments about current cases, according to the summary.
Another related project is investigating what is called a "Dynamic MRI Explorer" which is a prototype CAD workstation which can be used in the diagnosis of breast MRIs. The exhibit showcases the benefits of the workstation and also compares conventional MRI interpretation with the researcher's automatic algorithms. The study is being conducted by W. Chen, M. L. Ciger, G. Newstead, U. Bick, L. Lan.
Developments of the breast MRI CAD workstation are, as the research summary states:
- Fuzzy c-means clustering for automatic lesion segmentation and extraction of kinetic information from enhancing voxels in the lesion margin;
- Calculation from the MR image, including kinetic and spatial features for improved evaluation of a lesion; and
- An artificial neural network for auto merging extracted image features into an approximated chance of malignancy.
Last but not least is another presentation of a clinical CAD workstation for breast ultrasound. The system has been previously tested for ultrasound exams in retrospective studies which was successful. The exhibit now is focused on presenting the workstation to demonstrate planned future use within a clinical setting with "full output, including the display of computer analysis results," according to a summary.
The work with CAD workstations for breast ultrasound is being done by K. Drukker, N. Gruszauskas, C.A. Sennett, L. Lan, M. L. Giger, and I. Bonta, also a part of the Department of Radiology at the University of Chicago.