Worldwide, breast cancer is one of the most common cancer diagnoses among women.1 It is also one of the most studied cancers. A review of annual research spending in the U.S. shows more funding allocated for breast cancer research than for any other cancer type.2 The 34% drop in breast cancer mortality in the United States between 1990-20103 has been attributed to improvements in both detection and treatment, which reflects the high level of cancer research. The remaining prevalence of the disease, however, highlights the need for continued research and use of new technologies to drive breast cancer breakthroughs.1,3
Researchers across a variety of fields-including breast cancer-are increasingly using quantitative image analysis tools to support their methods. These tools allow them to measure biomarker data in a truly objective, quantitative fashion, which offers a number of advantages over manual qualitative or semi-quantitative techniques. This includes a generation of highly standardized and reproducible data, reduction of inter- and intra-observer variability and subjectivity. It also offers the ability to analyze histology images in a high-throughput fashion with minimal user interaction, reducing manual effort and improving turnaround time.
With the emergence of digital pathology, users now have access to a wide assortment of computer-assisted image analysis options-from basic pixel counting to highly specialized tools for specific applications.
Image Analysis in the Clinical Environment
Image analysis tools are capable of accurately quantifying biomarker staining in tissue, and they have the potential to act as a valuable diagnostic aid to the pathologists manually assessing samples. Numerous studies have investigated the potential for automated image analysis to move into daily clinical practice for breast cancer.
SEE ALSO: Cancer Diagnosis in a Drop of Blood
A multisite study of 260 breast tissue specimens, comparing quantification of ER and PR with automated image analysis, found substantial correlations between the automated and manual methods.4 The same group also evaluated HER2 automated analysis in the 260 specimen cohort and discovered that, not only were automated analysis results substantially equivalent to manual read, but the availability of quantitative data also improved interpathologist agreement.5
Fasanella et al.6 used a nuclear analysis algorithm to quantify Ki67 expression in 315 breast cancer samples previously evaluated by a pathologist, reporting a high level of correlation between manual and automated image analysis. Lloyd et al.7 compared ER and HER2 scoring by digital image analysis with both a manual read by two pathologists and the gold standard HER2 FISH, using a trained pattern recognition algorithm to identify tumor areas and nuclear and membrane analysis algorithms to quantify the biomarkers of interest within the tumor. They found that all of the image analysis results fell within acceptable range of a pathologist’s manual read.
Laurinavicius et al.8 used similar methodology, employing a pattern recognition algorithm in combination with nuclear and membrane analysis algorithms to quantify a panel of breast biomarkers, including HER2, ER, PR and Ki67. This study looked at the potential development of a multi-marker expression profile for Ductal Carcinoma and examined the use of image analysis “to obtain more accurate, reproducible and quantitative results.” Their conclusions described image analysis as, “An efficient exploratory tool clarifying complex interdependencies in the breast cancer carcinoma IHC profiles.”
Studies have also examined image analysis as a tool for quality control in the histopathology laboratory. For example, Laurinavicius et al.9 looked at use of digital image analysis for quality control of staining in clinical practice, using a pattern recognition algorithm to identify tumor regions and a membrane analysis algorithm to quantify HER2 expression within those tumor regions across different staining batches. They found that image analysis was able to detect staining drift that was not identified during conventional microscope review, making a potentially valuable tool for quality control of IHC staining.