AI Can Drive the Future of Cancer Pathology Soon

AI driven cancer pathology

Medical technology has been progressing immensely over the last decade. The kind of machinery and technology making its way into diagnoses and treatment has been phenomenal. But what has been relegated to the background is the field of pathology – which ironically – is the first step to definitive diagnoses of any kind.

Pathology has not lent itself easily to technological upgrades up until now. To enable the reading of slides in a digital format would mean converting glass to pixels – which means an immense amount of storage being required by diagnostic labs. This alone is an IT nightmare. Another issue to deal with is that the digitization of pathology will have to answer questions that so far have remained vague. Take cancer – patients and families would like to know the chances of full recovery, or re-occurrence, and of course possible severity. Human driven diagnoses has been limited so far in this respect. With digitization of pathology, there will be a need for medical innovations that will work in tandem to provide these answers.

All this may slowly come to reality now that the US FDA has approved the first whole slide imaging system for primary diagnostics. Also, Artificial Intelligence (AI) for digital pathology is increasingly being worked on.

So how can AI Help Pathologists Improve Cancer Care

AI in cancer pathology could possibly help in the following ways:

Pathology Images Can Be Better Interpreted by AI:

Machine-learning algorithms are currently being created that will better enable the interpretation of complex patterns that emerge in cancer pathology, based on real-life data that has been accumulated and created into artificial neural databases. These databases will bring in information from trillions of pixels worth of tissues that are scanned and have their information stored to enable better diagnoses.

Featured Whitepaper

healthcare

Enabling Early Diagnoses of Cancer:

Through the course of their careers, pathologists will have the ability to scan an average 10 million images of specimens and document what they have seen. AI on the other hand will have databases that have the ability to scan and analyse much more than this, quickly and accurately as well. The output that these machines are then able to achieve are phenomenal, making way for early stage diagnosis. This will help in something that is routine, but tedious and time consuming – like looking for malignant cancer cells among millions of normal ones.

Better Diagnoses of Lymphoma Subtypes:

Cancers can be of several different type. One study focussed on 3 main subtypes of lymphoma, which is a common cancer that has 50 distinct subtypes. Their research shows more than 5% to 15% of lymphoma cases have the chance of being misdiagnosed or misclassified initially, making treatment significantly difficult. With the right databases set and machine learning AI in place, this issue can be resolved and treatments made more accurate; many starting early as well. This can make the difference between life and death.

There is a lot that needs to happen before AI can enable pathologists to work better in the diagnosis and treatment plans related to cancer. One of the first steps is creating large datasets for the training of AI systems that will be used in pathology. Efforts are in progress and currently AI diagnostic abilities in one study stand at 92% accurate, going up to 99.5% when combined with a human pathologist. A human stands at 96% accurate.

The long term idea is to create techniques that analyse extremely high volumes of data quickly and accurately. The AI systems will work in tandem with humans, with idea of quicker diagnoses being the key. Doctor-friendly interfaces is what the focus will also need to be on. With the number of medical researches and technology companies working on this right now, and with the opening up of the scope of the industry thanks to the first approval from the FDA, there is a lot we can expect from AI for pathology.

Featured Whitepaper

healthcare wearable

Author : Ruth Date : 05 Dec 2017