HOW DEEP LEARNING IS TREATING HEALTH-BASED ISSUES

Rosie O'Callaghan our consultant managing the role
Posting date: 8/30/2018 8:00 AM
Hospitals are a complicated system of many moving parts both human and machine. In recent years, the role of humans driving the process, entering information, gathering individual records, or arranging medical and billing follow ups, has shifted.

Paper records have become electronic health records and AI is helping streamline bulky processes. AI bots and programs free up time when it comes to arranging follow up medication or helping to make diagnoses and, in some cases, can assist physicians or surgeons making remote calls and decisions.

As Machine Learning and AI enter healthcare, the application of Deep Learning, using data rather than task-based algorithms, is coming into its own. At this year’s KDD event, both Healthcare and Deep Learning were hot topics, with a day of programming dedicated to each.

The Three Ingredients Driving AI Advances:


  • Supply of digital data which can now be created.
  • Development of algorithms to make artificial neural networks.
  • Graphics Processing Unit (GPU) chip architecture pioneered by NVIDIA.

GPUs are used by anyone working in Deep Learning and can be used in any number of ways, such as videos, graphics, and audio recordings to name a few. This type of usage has huge impact on Healthcare’s image, clinical data interpretation, and management. 

For example, Radiology requires consultants to look at medical imagery to determine whether or not there are abnormalities. With the inclusion of Deep Learning, this process could be done in minutes or seconds rather than hours. This is especially important as a diagnosis made is based on findings in the radiological images.

However, Radiology, is not the only instance where health management can utilise Deep Learning and AI. From helping to identify ideal treatments for patients, to helping administrators utilise their resources more effectively and efficiently, there is huge potential for implementation. 

Predictive Analytics in Deep Learning


Healthcare can be hard to predict. But, with the application of Machine Learning, there are some things we can focus on, starting by asking ourselves the following:

  • Is it scalable? This may differ based on different hospital systems and how much data wrangling is involved. But, the more straightforward the answer, the better.
  • Is it accurate?  Using Deep Learning data for electronic health records can greatly improve accuracy and avoid the distraction of false alarms.

Predictive modelling can help Healthcare professionals answer the questions above more accurately, including determining which patient will have a particular outcome versus which patient will not. Though this model does not diagnose the patient, it does use the information from data gathered to identify the conditions in which the patient was being treated and predict outcomes. Like a human might pick up nonverbal signals, AI picks up signals based on the data it receives to and helps inform physician’s decisions.

The Patient Journey 


Whether it’s the customer journey or the patient journey, there is a path that needs to be followed. As Deep Learning helps fuel the use of AI in Healthcare, our patient journey becomes less stressful and more streamlined. 

Below are a few ways Deep Learning is helping to facilitate a more efficient health management system:

  • At Home: You go to a doctor because you don’t know what’s wrong. But, how do you know which doctor you should make an appointment with? AI can help. From your home PC, a few clicks and few questions can direct you to the correct provider for your needs.
  • In the Waiting Room: To avoid long wait times, you can check in via an app, have an AI bot ask a number of questions for you to answer to help better prepare the physician for your visit with the goal of a quicker diagnosis.
  • With the Doctor: Referrals are great. But, having to explain your health issue or record, can be daunting. In addition, the doctor to whom you’re referred may have to call your traditional physician and discuss, or he or she may have papers to read cutting into their time with you. Instead, AI standardises how the doctor reads the notes and can lay it out the way the doctor prefers, increasing your time with them and streamlining their process.
  • Patient Follow Up: An AI bot based on Deep Learning algorithms can become part of a provider’s team, checking in, asking a few questions, and sending a friendly reminder email, text, or phone call to remind patients to take continue their course of treatment. 

The introduction of Deep Learning into Data & Analytics has made an impact across many industries, but especially Healthcare. not the least of which has been healthcare. From speech recognition to Natural Language Processing, the effects have been informative and transformational.

If you’re interested in Deep Learning, predictive analytics, or AI we may have a role for you. We specialise in Junior and Senior roles.  To learn more, check out our vacancies. You can also call us at +44 20 8408 6070 or email us at info@harnham.com.

Related blog & news

With over 10 years experience working solely in the Data & Analytics sector our consultants are able to offer detailed insights into the industry.

Visit our Blogs & News portal or check out the related posts below.

WE HAVE TO TEACH SPECIALISATION, WE CAN’T EXPECT IT: A Q&A WITH VIN VASHISHTA

We recently spoke to Vin Vashishta, a consulting Data Scientist and Strategist who was named one of LinkedIn’s Top Voices in Data Science.  Having started off in the tech world 25 years ago and progressing from web design and hardware installation to Business Intelligence Analytics, Vin found for many years that enterprises were reluctant to adopt AI technologies and embrace the value of Data. In fact, it wasn’t until the beginning of the decade just passed that companies started to think about their Data more strategically and the world of Data Science was born, albeit hesitantly:  “When I first started, it was a lot of experimentation, everyone wanted a proof of concept,” he says. “A lot of work was creating models that could go from whiteboard to production and productise and show their value.” However, it wasn’t until halfway through the decade that he began to see businesses who had adopted Machine Learning move away from experimentation into incorporating it more deeply into their companies, relying more on analytical and optimisation models to make strategic business decisions.  “After that, in about 2017/2018 the maturity changed. It went from being a one off implementation to it being a comprehensive tool within an organisation where we have full lifecycles of model implementation and full models that were full views of the system. The key component of development was allowing users to access a small part of the system to do their job better without having to understand the whole thing. And that’s where we are now. We have this applied Deep Learning and we are seeing, especially this year, attempts to optimise that, make things go faster and make them more repeatable.” But, as we all know, with great power comes great responsibility: “There’s this whole depth we are getting into, the expectations are so much higher, people don’t just expect it to work they expect it to work the way they want it to and in a way they can adopt.” So, with so much expected and required of Data Scientists in 2020, building the right team is more important than ever. However, many businesses, Vin believes, are yet to get their hiring processes right: “A lot of the measures that we use to sort of evaluate employees are fictional – when you say years of experience, it has no correlation to employee outcomes or the quality of employee you get long term. It’s the same thing as college degree, there’s no correlation.” So when Vin is trying to build a highly specialised team, what does he do? “We have to teach specialisation, we can’t expect it. We can’t bring someone in and call them a Data Scientist and hope that they train up. You end up with teams that are exactly the same because they have hired the same people, people who reinforce the bias of what they do, and that is where true leadership needs to come in.” A specialised team made up of individuals who bring their own ideas to the table is more important than ever, particularly as businesses demand more from their Data teams. Gone are the days of one-size-fits-all models. Businesses now want something tailored to them: “Custom models are huge. The “import from…” Machine Learning development from three years ago adds value when it comes to wrangling and doing the Analysis, but when it comes to creating models companies are now expecting it to become a competitive advantage. Companies no longer want the same model that everyone else has, now it has to be differentiating.” These smart, customised models, he adds, will help businesses through the current pandemic. “The best models right now are adapting rather than reacting.”  However, he’s sceptical about the Data Science community becoming too preachy:  “When it comes to COVID-19 one message I want to send to the Machine Learning and Deep Learning community is ‘shut up’. We don’t have the Data! We have so many Data Scientists talking about something that’s very important to get right. If you get it wrong the consequences and the credibility we will lose as a field is enormous.” Indeed, discussions about the lack of quality Data on COVID-19 are widespread at the moment and raise concerns for Vin: “What the last two and a half months has revealed is the danger of bad Data, the danger of assumptions that are hidden in Data that hasn’t been looked over well or wasn’t gathered well and was fed into these models that now aren’t robust. Of course, no model can account for something this drastic, but they should still be performing far better than they are right now.” Despite these concerns, Vin believes any change in the world brings about opportunities for those in the Data and technology space. “What I’ve been trying to do ever since I joined the technology space is figure it out. It’s constantly evolving and it’s constantly changing. That’s really what has driven my journey. I’m always trying to figure out ‘what’s next’ over the next five years, ten years whatever it may be.” If you’re looking for your next Data Science, Machine Learning or Deep Learning role, or want to build out your own highly-specialised team, we may be able to help.  Take a look at our latest opportunities or get in touch with one of our expert consultants to find out more.   

The Search For Toilet Paper: A Q&A With The Data Society

We recently spoke Nisha Iyer, Head of Data Science, and Nupur Neti, a Data Scientist from Data Society.  Founded in 2014, Data Society consult and offer tailored Data Science training for businesses and organisations across the US. With an adaptable back-end model, they create training programs that are not only tailored when it comes to content, but also incorporate a company’s own Data to create real-life situations to work with.  However, recently they’ve been looking into another area: toilet paper.  Following mass, ill-informed, stock-piling as countries began to go into lockdown, toilet paper became one of a number of items that were suddenly unavailable. And, with a global pandemic declared, Data Society were one of a number of Data Science organisations who were looking to help anyway they could.  “When this Pandemic hit, we began thinking how could we help?” says Iyer. “There’s a lot of ways Data Scientists could get involved with this but our first thought was about how people were freaking out about toilet paper. That was the base of how we started, as kind of a joke. But then we realised we already had an app in place that could help.” The app in question began life as a project for the World Central Kitchen (WCK), a non-profit who help support communities after natural disasters occur.  With the need to go out and get nutritionally viable supplies upon arriving at a new location, WCK teams needed to know which local grocery stores had the most stock available.  “We were working with World Central Kitchen as a side project. What we built was an app that supposed to help locate resources during disasters. So we already had the base done.” The app in question allows the user to select their location and the products they are after. It then provides information on where you can get each item, and what their nutritional values are, with the aim of improving turnaround time for volunteers.  One of the original Data Scientists, Nupur Neti, explained how they built the platform: “We used a combination of R and Python to build the back-end processing and R Shiny to build the web application. We also included Google APIs that took your location and could find the closest store to you. Then, once you have the product and the sizes, we had an internal ranking algorithm which could rank the products selected based on optimisation, originally were based on nutritional value.”  The team figured that the same technology could help in the current situation, ranking based on stock levels rather than nutritional value. With an updated app, Iyer notes “People won’t have to go miles and stand in lines where they are not socially distancing. They’ll know to visit a local grocery store that does have what they need in stock, that they’ve probably not even thought of before.” However, creating an updated version presented its own challenges. Whereas the WCK app utilised static Data, this version has to rely on real-time Data. Unfortunately this isn’t as easy to come by, as Iyer knows too well:  “When we were building this for the nutrition app we reached out to groceries stores and got some responses for static Data. Now, we know there is real-time Data on stock levels because they’re scanning products in and out. Where is that inventory though? We don’t know.” After putting an article out asking for help finding live Data, crowdsourcing app OurStreets got in touch. They, like Data Society, were looking to help people find groceries in short supply. But, with a robust front and back-end in place, the app already live, and submissions flying in across the States, they were looking for a Data Science team who could make something of their findings.  “We have the opportunity,” says Iyer “to take the conceptual ideas behind our app and work with OurStreets robust framework to create a tool that could be used nationwide.” Before visiting a store, app users select what they are looking for. This allows them to check off what the store has against their expectations, as well as uploading a picture of what is available. They can also report on whether the store is effectively practising social distancing. Neti explains, that this Data holds lots of possibilities for their Data Science team: “Once we take their Data, our system will clean any submitted text using NLP and utilise image recognition on submitted pictures using Deep Learning. This quality Data, paired with the Social Distancing information, will allow us to gain better insights into how and what people are shopping for. We’ll then be able to look at trends, see what people are shopping for and where. Ultimately, it will also allow us to make recommendations as to where people should then go if they are looking for a product.”  In addition to crowdsourced information, Data Society are still keen to get their hands on any real-time Data that supermarkets have to offer. If you know where they could get their hands on it, you can get in touch with their team.  Outside of their current projects, Iyer remains optimistic for the world when it emerges from the current situation: “Things will return to normal. As dark a time as this is, I think it’s going to exemplify why people need to use Artificial Intelligence and Data Science more. If this type of app is publicised during the Coronavirus, maybe more people will understand the power of what Data and Data Science can do and more companies that are slow adaptors will see this and see how it could be helpful to their industry.”   If you want to make the world a better place using Data, we may have a role for you, including a number of remote opportunities. Or, if you’re looking to expand and build out your team with the best minds in Data, get in touch with one of expert consultants who will be able to advise on the best remote and long-term processes. 

RELATED Jobs

Salary

£75000 - £80000 per annum + Benefits

Location

Milton Keynes, Buckinghamshire

Description

B2C global company looking to build a brand new data science capability - this role will be the first hire!

Salary

€40000 - €45000 per annum

Location

Paris, Île-de-France

Description

This startup known in the insurance sector is looking for a Spanish Data Scientist to join their Data Science team in Paris.

Salary

US$180000 - US$200000 per year + Competitive Benefits

Location

San Francisco, California

Description

Harnham is working with a massive late-stage venture that is paving the way for machine learning. Let's talk about deep learning and/or ops research!

recently viewed jobs