Doctors turning to data to speed up diagnosis

Medical professionals are tapping into a wealth of online data in a bid to speed up the diagnosing of patients.

When you go into your doctor's office, it's likely they will have 15 minutes to assess, diagnose and provide a treatment plan for you before moving on to the next patient.

Doctors are well trained experts in their field who have to process new patient information, make a conclusion and take a course of action in a relatively short period of time. All of this expertise comes from their experiential knowledge based on cases they have studied or seen before.

What your doctor can't do is read the equivalent of a million books a second or follow the 170,000 clinical trials that are being carried out worldwide. This is where high-powered computing, using insights from machine learning, comes in for a new type of technology-enabled healthcare.

Machine learning is computer software designed to build complex mathematical structures called neural networks. These network software simulations mimic the way our brains connect lots of densely interconnected data and learn from it. Machine learning isn't capable of 'thinking' like a human, instead it uses computer processing power to quickly and precisely move through, file and store data.

Ninety per cent of the world's data was created over the last two years and it's estimated that we now produce 2.5 quintillion bytes of data a day. This is the world of 'big data', a phrase used to describe a massive volume of structured and unstructured data that is so large it's difficult to process using traditional databases.

Collecting the data is relatively easy; sifting through it all to make sense of the relevant bits is the difficult part and this is where machine learning comes in.

If you've ever made a typo while searching on Google, the "Did you mean..." suggestion often pops up. This is basic machine learning in action. Google's algorithm recognises when users type one thing, and then search for something similar a couple of seconds later.

It keeps this information and uses it to help future users who make a similar typing mistake. Essentially it's learning from peoples mistakes and suggesting corrections based on this learned information.

In the US, IBM and Quest Diagnostics recently launched IBM Watson for Genomics - a service which analyses cancerous tumours to determine the best course of treatment for an individual.

Tumour mutations vary between patients, and the success of a treatment depends on how well a doctor can match a therapy to a mutation. By sequencing the genomic makeup from a biopsy of the tumour and feeding the genetic files into IBM Watson, the computer can compare the mutations to a huge database of clinical studies, medical literature and work from leading oncologists, to determine the best treatment for that individual patient.

In New Zealand a new initiative called Precision Driven Health brings together Orion Health's collection of more than 100 million patient health records with the expertise of the Waitemata District Health Board and The University of Auckland.

By inputting a person's unique characteristics, including health and well-being, clinical and genetic information and adding environmental and lifestyle factors, Precision Driven Health applies machine learning so medical treatments can be precisely tailored to suit the patient's needs.

Initially designed to help us solve number crunching problems, toady's computers combined with machine learning are now being used to help healthcare and many other pattern-based fields, including fraud detection, speech understanding and facial recognition.

Thanks to the major software companies now making these complex systems much easier for non-experts to use, what was once reserved for highly technical computer coders is now opening up the power of big data and its medical advice to us all.

Dr Michelle Dickinson, also known as Nanogirl, is an Auckland University nanotechnologist who is passionate about getting Kiwis hooked on science. Tweet her your science questions @medickinson.