Source: Deep Learning on Medium
Deep Learning in Digital Pathology : a simple introduction
Expert in software architectures and passionate about deep learning in medical image
The actual world is telling us that that there are few experts in digital pathology in the world, their availability is very restricted, long education and training are barriers for learn skill in certain fields. One of this field is medicine, few university and a strict selection for enter the market are the firsts steps that identify/distinguish the process for become a doctor.
How to fight the lack of knowledge in the entire world, how to share the knowledge, the priceless expertise that could save humans life ? Artificial intelligence, a tool that could replicate mechanical steps from experts could help.
What can do for us deep learning for digital pathology?
We can have a look first and understand what is a pathologist.
What is a pathologist?
A pathologist is a medical healthcare provider who examines bodies and body tissues, trough lab tests, tools, diagnostics instruments. A pathologist helps other healthcare providers reach diagnoses and could propose treatment based on historical data and use-cases.
Digital pathology is an image-based information environment which is enabled by computer technology that allows for the management of information generated from a digital slide. Digital pathology is enabled in part by virtual microscopy, which is the practice of converting glass slides into digital slides that can be viewed, managed, shared and analyzed on a computer monitor. With the advent of Whole-Slide Imaging, the field of digital pathology has exploded and is currently regarded as one of the most promising avenues of diagnostic medicine in order to achieve even better, faster and cheaper diagnosis, prognosis and prediction of cancer and other important diseases.
Normally we have images of tissues, coming from biopsy, where are present cancer cells that must be analyzed by pathologist, that will recognize which kind of tipology, stage and prediction phase of the cancer that the concerned tissue is affected.
This is taking several hours from an expert pathologist, that could take 10–12 years of training to reach the level. Can you imagine the effort to spend 6–8–10 hours drawing circles in a screen for identify tumor cells one-by-one? Imagine the fact that is not only matter of time but expertise to be able identify exactly the “bad” cells.
This tedious and mechanical way to do can be substitute by an “Artificial Intelligence” model that can be trained with the existing annotations of pathologists that has been done this for decades, nowadays.
This will not replace the pathologist himself, will give him a powerful tools for help, compare his insight and save times that can be spent in others activities, like research or clinical visits.
It’s important to explain what “Artificial Intelligence” mean for this specific case, doing so the concept of Deep Learning, a son of the family of the Machine learning should be explained.
Saying that Machine learning (ML) is the study of algorithms and statistical models that computer systems use to perform a specific task. This could be resumed as different repetitive tasks could be done automatically and the results could be forecasts, specifying that those actions should be observed for a very huge amount of times. In this deep learning is a technique that identify the results of those steps trough different phases, recently used in different fields like computer vision, speech recognition, natural language processing, audio recognition, social network filtering, machine translation, bioinformatics, drug design.
Resuming if you know the concept of image recognition of pictures ( like the facial recognition ), this could be transposed to the image recognition of tumor cells instead of faces; the image processing for identify tumor cells.
So now that you can catch the concept that tumor cells could be recognized in a microscopy image like people could be identified in a Facebook picture by faces, you have the big picture and you can assume in big lines what the mechanism can produce and can work.
The problem that can solved is to solve the vacancy of pathologist and can reuse the acquired knowledge of retired or dead pathologists
As a human, your brain can approach most any situation and learn how to deal with that situation without any explicit instructions. If you sell houses for a long time, you will instinctively have a “feel” for the right price for a house, the best way to market that house, the kind of client who would be interested, etc. The goal of Strong AI research is to be able to replicate this ability with computers.