Personalized breast cancer treatment based on Sox2 protein expression

Objectives: Breast cancer is the most common cancer type among women worldwide. However, it is a very heterogenous disease and in many cases, resistance to standard endocrine therapy develops. Personalized medicine, the approach that proposes to tailor the treatment process to the individual needs of each patient, is recently gaining increasing attention in the development of effective treatments. Recent studies have identified Sox2, a transcription factor that is key in maintaining pluripotent properties of stem cells, as a crucial player in the development of resistance to endocrine therapy in ER-positive breast cancer. Therefore, Sox2 protein may represent a biomarker to predict treatment resistance. The analysis of Sox2 protein by immunohistochemistry (IHC) staining can enable us to identify patients with the highest risk to be able to decide on the most appropriate therapy. However, the manual assessment of the images is cumbersome and quantifying the percentage of IHC staining can be a challenge for pathologists. In this study, we have developed an automatic method to identify, classify and quantify Sox2 expression. Methods: The measures are extracted at pixel and nuclei levels using a K-means algorithm for segmentation and a colour filtering approach based on channels (RGB, HSV) to separate Sox2 positively and negatively expressed nuclei. Results: 75 images were analysed using our method and compared to pathologist scores. Preliminary results demonstrate that the proposed method has high precision (94%) and recall values (93%). Conclusions: This method measures Sox2 expression in an objective, accurate and reproducible way that will help clinicians to decide what treatment will be best for the individual patient, delivering personalized medicine and enhancing the effectiveness of currently available cancer therapies.