Improving Accuracy of Nuclei Segmentation by Reducing Histological Image Variability

Cancer is the second leading cause of death in United States. Early diagnosis of this disease is essential for many types of treatment. Cancer is most accurately observed by pathologists using tissue biopsy. In the past, evaluation of tissue samples was done manually, but to improve efficiency and ensure consistent quality, there has been a push to evaluate these algorithmically. One important task in histological analysis is the segmentation and evaluation of nuclei. Nuclear morphology is important to understand the grade and progression of cancer. Convolutional neural networks (CNN) were used to segment train models for nuclei segmentation. Stains are used to highlight cellular features. However, there is significant variability in imaging of stained slides due to differences in stain, slide preparation and slide storage. This make automated methods challenging to implement across different datasets. This paper evaluates four stain normalization methods to reduce the variability between slides. Nuclear segmentation accuracy was evaluated for each normalized method. Baseline segmentation accuracy was improved by more than 50% of its base value as measured by the AUC and Recall. We believe this is the first study to look at the impact of four stain normalization approaches (histogram equalization, Reinhart, Macenko, Khan) on segmentation accuracy.