This article is more than 5 years old. Information may no longer be current.
Imaging systems may diagnose breast cancer without histologic assessment
Click Here to Manage Email Alerts
Proflavine staining and confocal fluorescence microscopy in combination with image analysis strategies to segment morphological features may diagnose breast cancer in fresh breast tissue without a pathologist, according to study results.
“Breast cancer diagnosis is a complex process, which requires tissue procurement through core needle biopsy or surgical excision, followed by rigorous tissue preparation and histologic assessment,” Rebecca Richards-Kortum, PhD, Stanley C. Moore professor of bioengineering and professor of electrical and computer engineering at Rice University, told HemOnc Today. “In many parts of the world, a lack of economic resources or trained personnel makes evaluating suspicious breast masses a very expensive and time-consuming process. Our aim with the imaging techniques in this study was to potentially improve access to histologic diagnosis in the many parts of the world that lack the human resources and equipment necessary to perform standard diagnosis of breast masses.”
Rebecca Richards-Kortum
Current breast cancer diagnoses are made through histologic assessment, which requires fixation and tissue preparation by a pathologist.
Diagnostic criteria are qualitative and subjective, according to the researchers, with inter-observer discordance posing a particular challenge in the diagnosis of selected breast lesions.
Richards-Kortum and colleagues acquired images of freshly excised breast tissue specimens (n = 259 sites) from a total of 34 patients using confocal fluorescence microscopy and proflavine as a topical stain. They then developed computerized algorithms to segment and quantify nuclear and ductal parameters that characterize breast architectural features.
The researchers evaluated 33 parameters and used them as input to develop a decision tree model to classify benign and malignant breast tissue. They classified benign features in tissue specimens acquired from 30 patients (n = 179 sites) and malignant features in specimens from 22 patients (n = 80 sites).
The decision tree model that achieved the highest accuracy for distinguishing between benign and malignant lesions included standard deviation of inter-nuclear distance and number of duct lumens. The model made diagnoses with 81% sensitivity and 93% specificity, with an area under the curve of 0.93 and 90% overall accuracy.
Further, the model diagnosed invasive ductal carcinoma with 92% accuracy and ductal carcinoma in situ with 96% accuracy.
A cross-validated model achieved 75% sensitivity and 93% specificity, with an overall accuracy rate of 88%.
The researchers acknowledged limitations of their study, including the small number of patients and single-institution study design. Further, they noted that the high cost, footprint and maintenance requirements of confocal microscopy may preclude its widespread routine usage.
“The findings from our study demonstrate that quantitative diagnostic criteria have the potential to enable automated assessment of fluorescence confocal images of breast tissue,” Richards-Kortum said. “The parameters we describe based on breast ducts and nuclei could be used to develop criteria to immediately evaluate fresh tissue at the point of care with minimal tissue preparation. Further, our approach could reduce subjectivity between physicians.” – by Cameron Kelsall
For more information:
Rebecca Richards-Kortum, PhD, can be reached at Rice University, Bioengineering, MS-142, PO Box 1892, Houston, TX 77251-1892; email: rkortum@rice.edu.
Disclosure: Richards-Kortum reports an advisory role with and ownership interest in Ramicalm, LLC. One other researcher reports a founding role in Zenalux Biomedical, which has developed technologies related to this study.
Click Here to Manage Email Alerts