Grey Level Texture Features for Segmentation of Chromogenic Dye RNAscope From Breast Cancer Tissue
This work addresses the need for automated quantification methods in pathology workflows for breast cancer diagnosis, though it is incremental as it applies existing texture features to a new application.
The paper tackled the problem of automating the segmentation and classification of RNAscope transcripts in breast cancer tissue, which is time-consuming manually, by using grey level texture features and achieved an F1-score of 0.571, similar to expert inter-rater performance of 0.596.
Chromogenic RNAscope dye and haematoxylin staining of cancer tissue facilitates diagnosis of the cancer type and subsequent treatment, and fits well into existing pathology workflows. However, manual quantification of the RNAscope transcripts (dots), which signify gene expression, is prohibitively time consuming. In addition, there is a lack of verified supporting methods for quantification and analysis. This paper investigates the usefulness of grey level texture features for automatically segmenting and classifying the positions of RNAscope transcripts from breast cancer tissue. Feature analysis showed that a small set of grey level features, including Grey Level Dependence Matrix and Neighbouring Grey Tone Difference Matrix features, were well suited for the task. The automated method performed similarly to expert annotators at identifying the positions of RNAscope transcripts, with an F1-score of 0.571 compared to the expert inter-rater F1-score of 0.596. These results demonstrate the potential of grey level texture features for automated quantification of RNAscope in the pathology workflow.