The project builds on IBM’s pre-existing agreement with MoleMap – a company that uses advanced visual analytics to examine over 40,000 data sets, including image and text. IBM wants to look at dermatological images of skin lesions to identify specific patterns in the early stages of melanoma.
A joint IBM MoleMap investigation in 2015 resulted in IBM Research performing retrospective analysis of complex and versatile data sets of 40,000 images, dermatology opinions, and diagnoses across three types of skin cancer and 12 benign disease groups. This initial formative testing indicated an accurate detection of melanoma in 91% of all dermascopy images and 83% on clinical photography images.
“Melanomas are often missed in routine skin checks, as doctors don’t always have the technology or skills to recognize them, particularly in the early stages. Our aim at MoleMap is to detect melanoma earlier and more accurately, so partnering with IBM and using its cognitive capabilities to help drive this forward made perfect sense for us. We’re looking forward to seeing what we can achieve together to help fight this deadly disease,” said Adrian Bowling, MoleMap’s CEO
Because melanoma and skin cancer are a high priority for Australians, the project is aiming to contribute to the reduction of unnecessary biopsies, help clinicians improve their understanding of skin cancer, and improve patient care. Current statistics suggest that two in every three Australians will be diagnosed with skin cancer before turning 70 – even though 90% to 95% of all skin cancers are preventable.
Early diagnosis of skin cancer is crucial for sustaining positive survival rates, especially for melanoma, which is considered one of the most dangerous types of skin cancer.
In a press release, Melanoma Institute Australia’s Research Director Graham Mann said the project will eventually save lives.
“The five-year survival rate for melanoma is only 64% once the disease reaches the lymph nodes. However this rises to 95% if detected before then. Diagnosing melanoma with the naked eye is only about 60% accurate, but dermascopy can lift that to over 80%. Research using automated analysis of images could provide the next gain in accuracy, especially where dermascopy is hard to access,” Mann said.
Melanoma Institute Australia boasts the broadest melanoma research and treatment facility and the largest melanoma research database in the world.
IBM will use advanced visual analytics to conduct retrospective analysis. The project will comprise over one million images from 9,000 Australian and New Zealand patients, as well as clinical notes, to improve the accuracy of the machine learning algorithms. IBM is developing a type of cognitive technology that aims to learn how to understand skin cancers like melanoma, basal cell carcinoma or squamous cell carcinoma through the use of lower resolution clinical images. The goal is be as accurate as dermascopy images.
“This initiative could inform future research and, potentially, the development of offerings that could have enormous implications for both the Australian public and the health system,” said Dr. Joanna Batstone, IBM Research’s vice president and lab director.