From the course: Leveraging Cloud-Based Machine Learning on Azure: Real-World Applications

Healthcare - Azure Tutorial

From the course: Leveraging Cloud-Based Machine Learning on Azure: Real-World Applications

Start my 1-month free trial

Healthcare

“

- So let's take a look at a machine learning use case and we'll start with a medical machine learning based application. So we have a couple of requirements to consider. Number one the input. Number two the model and how we're going to process the data that's inputed. And number three the output of the information. So the input's going to be patient data and treatment data and that's typically going to be tagged information and so it's going to come into the machine learning system as an input to the knowledge model. So we understand we may have petabytes of patient data that could span tens of thousands of patients. And we could have hundreds of thousands of treatment data and that's typically going to be one to many granularity carnality with patient data to treatment data. When building the model we need to figure out there's patterns of success, in other words successful outcomes of the treatment data to the patients. And patterns of failure, and the patient got worse or worst things happen. And ultimately the ability to come up with a common diagnostics and the ability to apply the diagnostics to get to a likely outcome. So we're able to take the patterns of success, we're able to consume that information into our machine learning system. Able to discern the diagnostics are going to lead to likely success. We want to repeat that over and over again. So we're able to leverage data ingestion and this time we're taking an unsupervised machine learning approach and so the data is going to be clustered and not tagged. We're automating the movements and the duration of the meta data and we're able to discern actions to outcomes. In other words, what's a good outcome, what's a negative outcome based on the treatment data that's applied to a patient's symptoms and the patient themselves. And the ability to have that clinically delivered to various systems, either the doctors and the nurses who are treating the patients, or perhaps the surgeons who are performing surgery on the patients. And the idea is to leverage this data in a meaningful way so that we're able to apply correct treatments based on what we see within a particular patient based on the loads of experience that we're referencing through the data input into our machine learning system. So the results are early detection, the ability to spot problems early. Outcome based learning is going to be ongoing. So in other words we're able to pull the information back into the system so it can be re diagnosed by a machine learning system, and therefore better the outcomes as it learns more. And in this case we've had a 40% increase in survivorship because we're able to diagnose issues based on the massive amounts of data that's filling up our knowledge engines, that allows us to discern successful outcomes based on the information we have.

Contents