Fighting Skin Cancer with Visual Analytics

With 50,000km of coastline, Australia has some of the most beautiful beaches in the world.  And with our love of sport and outdoor adventure, it comes as no surprise that Australia also has one of the highest incidences of skin cancer in the world – two to three times the rates in Canada, the US and the UK.

1 in 2 Australian men and 1 in 3 Australian women will be diagnosed with cancer by the age of 85, which brings not only a significant loss of lives each year, but also costs the economy more than $3.8 billion in direct health system expense.

The sooner a skin cancer is identified and treated, the better chance of avoiding surgery and/or preventing death.  Fortunately, humans are pretty good at identifying positive and negative cases of skin cancer, with the highest level of accuracy achieved at 84%.  But imagine if we could increase the accuracy to more than 95%?  Given Australian GPs are faced with over 1 million patient consultations each year for skin cancer, that could translate into 110,000 more accurate diagnoses each year.

You can see why I’m really excited about the project that IBM Research and Memorial Sloan Kettering Cancer Center are currently working on – using visual analytics to increase the accuracy of skin cancer diagnosis.

Using cognitive visual capabilities being developed in the IBM research labs, computers are consuming vast amounts of educational research data, and being trained to identify specific patterns in images and perform finely detailed measurements that would otherwise be too large and time consuming for a doctor to perform.  This includes analysing the objective quantification of visual features – color distributions, texture patterns, shape and edge information etc.

Algorithms are also used to measure the progression of lesions, such as aggressive growth over a short period of time, or deviations from what is considered “normal” for a specific patient or population.

This analysis will be provided to clinicians to highlight dermatological images that may signify disease.

In preliminary trials of over 3,000 cases of melanoma, atypical lesions and benign lesions, the technology developed by IBM recognised diseased states with 97% sensitivity, and is generating significant excitement about the future impact of such technology being widely adopted in the industry.

Once commercialised, the cognitive computing approach will be able to scan the images in less than a second, much more quickly than humans can, and provide insight to doctors so they can make more informed decisions about how to progress.

There is still much to be done.  IBM Research will continue to work with Memorial Sloan Kettering to further refine and develop the process through larger data sets, but the initial results are extremely exciting and I can’t wait to see this technology make it to Australian shores.

 

IBM Visual Analytics to Fight Skin Cancer