COMPUTER-AIDED TOOL FOR DIAGNOSTIC BREAST ULTRASOUND

DESCRIPTION (provided by applicant):
Almen Laboratories has developed a sophisticated software system for imaging applications that provides extensive tools to identify objects and image features of interest, analyze the information content and then store, retrieve and compare different objects and images of interest based on this information. This proposal adapts the technology to the problem of breast ultrasound Level of Suspicion (LOS) computer-aided scoring (under control of practitioner) with the goal to reduce unnecessary biopsies. The aim of this program is to enhance the developed software tool to score LOS for cancer following the established BIRADS lexicon criteria using a form of case-based reasoning. Phase I will concentrate on demonstrating this technology in the clinical cases involving digitally acquired masses of the spectrum of classes: cystic, solid benign and carcinoma. This initial effort will be aimed at enhancing and optimizing the tool to accurately identify masses (target - 92% and above) with lower levels of suspicion, rather than increasing the accuracy of diagnosis of cancers in highly suspicious masses, which require biopsy anyway.

The system compares a breast mass in question to a database of images with known findings, displays those closest in Relative Similarity and computes an estimate of LOS with the physician in the loop. We hypothesize that the specificity of interpretation of breast ultrasound can be significantly improved with no significant change in sensitivity computer-aided imaging system to implement the structured BIRADS method for describing and scoring LOS for cancer. Lesions of lower LOS such as complex cystic masses can be ruled out as candidates for biopsy with higher degree of confidence with the computer-aided imaging system. The subsequent phases of the research will include application and comparison of different computer aided classification methods, issues related to machine dependencies, and adaptation of other patient risk factors such as mammography and physical exam findings. When these later phases are completed the system will be tuned to diagnostic breast ultrasound with focus on detection of missed cancers.

The market-ready application will be incorporated into the ultrasound instrument or workstation such that image scoring and classification is available on-line and/or in real time during scanning and biopsy procedures. A long-term goal of this program of research is to use the computer-aided classification system to create a novel database to archive, retrieve and compare medical objects of interest based on the their content. This database may serve as a valuable teaching resource, as an expert "second reader" resource to support diagnosis of suspicious unknowns, and as a model to design highly efficient computer networks for radiology departments through integration into the existing PACS systems. The tool has potential application to a wide variety of medical imaging problems including monitoring tumor response to therapy, classification of arterial stenosis, assessment of osteoporosis, assessment of brain function, three-dimensional visualization, etc. The open architecture design and multi-modality capabilities enable the system to be embedded in advanced PACS.