At Infosys, our focus on Healthcare is aimed at radical progress in affordability, wellness, and patient-centricity. We believe technology is a catalyst for game-changing healthcare solutions. In this blog, we discuss challenges, ideas, innovations, and solutions for the healthcare economy.

« What's next for Health IT - Trends for 2018 | Main

Deep learning on cloud for the healthcare industry

When you are not feeling very well, you would go to see a doctor hoping to get urgent medical attention and you would probably be the one person who can explain the symptoms to the doctor that even your loved ones cannot do accurately. The doctors need your medical history, current symptoms, other medicine regiments that you are probably on, etc. This information the patient provides is helpful in diagnosing the condition especially for typical cases like flue, type 2 diabetes high blood pressure, etc. The doctors have nearly adequate tools like the blood test, ECG, EEG and others to help them diagnose much more complex issues. However, with all the medical images, data and history, a significant amount of complex cases get attributed to idiopathic in nature. That means, the cause of the issue is unknown and the diagnosis of the issue is not straightforward. Brain tumors for instance; the doctors would have a good amount of MRI scanned images, radiology and EEG graphs, but still, end up with a considerably lower confidence diagnosis of the issue. In some extreme cases, invasive procedures are carried out just to get more information on the condition. You cannot expect a doctor or a team of doctors to have seen the exact complex situation earlier that can give them a 100% confidence on the matter at hand.

Off late, Artificial Intelligence and Machine Learning frameworks are proving to be of great help in assisting the doctors to diagnose many such complex cases. Though comparatively new, the machine learning software can help in enabling computer programs to progressively learn from already diagnosed and confirmed scan images in great detail and be able to predict and signal the doctors to take appropriate steps in determining the cause of the ailment. These images need not contain other patient's personal information at all. They come from all around the world, not just one health care entity, scrutinized for masking all the patient's personally identifiable information, and then fed to these artificial intelligence programs to learn in great detail as to which condition can result in what all different sort of MRI images and carefully classify the new images into false positives or potential threatening conditions. Now, the doctors are able to not just look at the raw MRI images, but also have hints from the computer program that has learnt to identify such cases over and over again. This new image fed would also become a part of the so-called 'brain' of the AI software to help it make future predictions in a more confident manner.


Back in 2016 itself, IBM Watson announced ground-breaking results by matching at a staggering 90% consensus with the tumor board's manual recommendations in a double-blinded validation study conducted by a team of oncologists at San Antonio Breast Cancer Symposium. Arterys has extended the same concept of using classifiers with supervised deep learning in helping Cardio, Lung and Liver diseases and indications. The Arterys set of AI imaging went even a step further and got FDA approval for clinical application! Google's DeepMind health is using retinal scan images to help provide more detailed data points for the treatment of macular degeneration in eyes which affects nearly 40% of the people over age 70. The number of applications of AI and ML in healthcare is growing at a fast pace and looks very promising.


Given the nature of the high compute power and resource intensive image comparison, IBM, Amazon Web Services, Google Cloud, Azure and many more cloud vendors have come forward to help implementation of such AI models and run them on the cloud. They can help train and retain the models on cloud and also provide an easy interface to access and use the intelligent programs and apply them to help every doctor around the world to make faster, better and more confident diagnosis to help the patients as efficiently as possible.


Most of the leading cloud providers have a strong Machine Learning (ML) and Artificial Intelligence (AI) agenda. Google's DeepMind and HyperTune ML engine helps developers tune their ML model training to optimize the processing. IBM's Watson ML Bluemix service offers a broader set of ML tools for prediction coupled with easy RESTful access. Microsoft Azure's Studio, on the other hand, provides a simple browser-based, drag and drop authoring environment for developers to explore. AWS Machine Learning engine has also been around for quite a while and is actively used by multiple vendors including Netflix. In fact, aside from the healthcare use case, application of ML in Netflix is pretty amazing. Each Netflix user sees a personalized recommendation of what to watch next with different title icons in the catalogue as well. What icon I see for Breaking Bad and Better Call Saul, as a crime series lover, is different than what you as a Thriller genre lover sees for the same two shows. This level of customization by Netflix per user has been possible by the Amazon ML engine they use underneath! This very use case shows the maturity and readiness of these cloud vendors to be able to handle a huge amount of data and still be computationally efficient in processing complex algorithms, images and sustain the models for practical usage.


This is a big step forward for Applied Artificial Intelligence and Machine Learning in the healthcare industry. Combine this technology with the cloud to power the accessibility, we have a winner combination. The day is not far when every doctor can have the preliminary diagnosis presented to them by a medical AI assist-bot and the doctor adds his experience as well to it to make the right call for the treatment.

Post a comment

(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)

Please key in the two words you see in the box to validate your identity as an authentic user and reduce spam.

Subscribe to this blog's feed

+1 and Like

Follow us on

Blogger Profiles

Infosys on Twitter