AI in Computational Pathology
The article aims to give a gentle introduction to Computational Pathology and the applications of AI in Computational Pathology.
We will be starting with an Introduction and then gradually move to AI in Computational Pathology.
Cancer
It has been reported by WHO that Cancer is the second leading cause of death annually with 9.8 Million deaths in 2018. Cancer occurs when normal cells are transformed into tumor cells. This behavior can be either caused by Genetic or aging factors.
Cancer Diagnosis
We now know that cancer is a fatal disease, therefore cancer diagnosis, typically early cancer diagnosis, is crucial. There are various ways in which cancer can be diagnosed and a few of these methods are:
- PET Scan
- CT Scan
- Genomic Profiles
- Tissue Analysis (Computational Pathology)
The focus of this article would be on cancer diagnosis using Computational Pathology.
Cancer Staging
Cancer Staging in simple words can be said as differentiation of Aggressive Cancer from Non-Aggressive Cancer. Differentiation (Staging) helps in inferring the growth and stage of the tumor. This eventually helps the pathologist in planning treatment for the patient which can be Radiation Therapy, Chemotherapy, Immunotherapy, etc.
Whole Slide Image (WSI)
Cancer Diagnosis and Cancer Staging can only be carried out with the help of the Whole Slide Images(WSIs). Therefore it is very important to know about WSI and its acquisition.
Digitized Tissue Slides are known as Whole Slide Image(WSI) these WSIs comprise of Multi Giga Pixels. The digitization process steps comprise of:
- Extraction of Biopsy Sample from Patients Body
- Placing Tissue Samples on Glass Slides
- Staining of Glass Slides
- Scanning of Glass Slides via Digital Slide Scanner
The scanning of Glass Slide is done in a manner that generated WSI imitates the behavior of a microscope i.e the WSI has multiple zoom levels ranging from 5x to 40x depending upon the type of scanner used.
There are various image file formats in which WSIs are stored however the most commonly used format is TIFF. Here is an extensive list of file formats in which WSIs can be stored along with their attributes.
Computational Pathology Workflow
The Computational Pathology Workflow comprises Data Acquisition followed by Image Analysis, Computer-Aided Diagnosis, and Survival Study. However, these steps may vary in different studies.
Data Acquisition
The data acquisition process comprises various steps. Initially, a biopsy sample is extracted from a patient's body which is placed onto a glass slide after that it is stained using various staining methods to make the tissues visible to the naked eye. Next, the stained tissue glass slide is digitized with the help of Whole Slide Image Scanner which converts glass tissue slides into 2D Digital Images that comprise of Multi Giga-Pixels.
Image Analysis
Image Analytics comprises of identifying (Segmentation and Classification of Nuclei) Cell of various types from Tumor Microenvironment depending on the type of cancer we are performing the diagnosis for. Once identification of cells is complete, next comes the part of Identification of Digital Biomarkers which can vary for each type of cancer. However, some of the digital biomarkers can be cell counting, cellular community detection, presence of lymphocytes around tumor cells, etc.
Computer-Aided Diagnosis
Once the tissue image analytics part is complete Computer-Aided Diagnosis (CAD)is carried out. The goal of CAD is to assist a pathologist in taking decisions while carrying out various pathology tasks as well as helping in determining the stage and type of disease. The tasks in CAD can be disease assessment, disease grading, patient treatment planning, precision medicine, etc.
Survival Study
Survival Study uses Image Analysis and Computer-Aided Diagnosis for predicting the survival of a given patient. In Survival Study time to event analysis is carried out in which death of the given patient or survival time of a patient for a certain treatment is to be predicted. To make this happen various regression-based survival analysis models are used which namely are Cox Proportional Hazards Model and Kaplan Meier Curve etc.
About Syed Nauyan Rashid
- Linkedin Profile: https://www.linkedin.com/in/nauyan/
- Github Profile: https://github.com/nauyan
About Red Buffer
- Website — redbuffer.net
- Linkedin Page — linkedin.com/company/red-buffer