Imaging software being developed at Rice University in Houston could offer fast, accurate diagnosis of breast cancer without the need for a specialist, according to a study published in Breast Cancer Research.

The software uses quantitative diagnostic criteria to perform an automated histological assessment of breast cancer from tissue samples in a process that doesn’t require fixation and tissue preparation.

“This cuts out the tissue preparation process and allows for rapid diagnosis,” Rebecca Richards-Kortum, PhD, senior author and professor of bioengineering and of electrical and computer engineering, said in a statement. “It is also reliant on measurable criteria, which could reduce subjectivity in the evaluation of breast histology.”

To create the software, Richards-Kortum and colleagues used images of freshly excised breast tissue specimens from 34 patients using confocal fluorescence microsopy. Proflavine was used as a topical stain.

Algorithms were used to segment and quantify nuclear and ductal parameters, with a total of 33 parameters evaluated and used in the development of a decision tree to classify benign and malignant breast tissue.

Standard deviation of inter-nuclear distance and the number of duct lumens were the parameters that helped the decision treat model achieve the highest accuracy for distinguishing features. This optimized model had an overall accuracy of 90 percent (81 percent sensitivity and 93 percent specificity).

The model classified invasive ductal cancer with 92 percent accuracy and ductal carcinoma in situ with 96 percent accuracy.

“To evaluate fresh breast tissue at the point of care could change the current practice of pathology,” said Richards-Kortum. This would be a particular boon in developing countries with fewer specialists.

The researchers did note that confocal microscopy is currently costly, limiting the ability to translate the imaging-based technique into routine usage for the time being.