Pathology plays a crucial role to diagnose cancer and to assess its progression from hematoxylin and eosin (H&E)-stained tissue samples. Diagnosis and assessment of tissue samples on glass slides have been done under microscopes which can be inefficient and subjective. Digitization of glass slides and deep learning-based computational approaches have been investigated to help this process. Especially, semantic segmentation, also known as pixel-wise classification, of whole slide images providing information of location and size of multiple tissue subtypes is a prerequisite step for clinical interpretations. In this talk, I will present how tissue segmentation can be done using deep learning and assist cancer diagnosis and assessment in an efficient and reproducible manner. Specifically, (1) I will describe Deep Multi-Magnification Network for multi-class tissue segmentation which looks at morphological features from multiple magnifications for more accurate segmentation. (2) I will explain Deep Interactive Learning to efficiently annotate whole slide images to train segmentation models. (3) I will introduce three applications where pathologists can be assisted by automated tissue segmentation. A breast model segmenting cancer can highlight cancer regions and screen potential malignant margin slides for efficient diagnosis with high sensitivity where margins for lumpectomy are mostly benign. An osteosarcoma model segmenting viable tumor and necrotic tumor can reproducibly estimate case-level necrosis ratio from multiple slides for pre-operative treatment response assessment which is used for prognosis. A lung model segmenting multiple tumor subtypes can find the predominant pattern and objectively determine cancer grading. In conclusion, tissue segmentation models can provide efficient and objective supports to pathologists.
Posted by: Kris Campbell