The technology has been devised at the University of Waterloo, together with a team from the Sunnybrook Research Institute. The focus is with the detection of detect melanoma skin cancer. The technology utilizes machine-learning software in order to analyze images of skin lesions. The analysis seeks to provide doctors with objective data on biological markers of melanoma. This is important since early detection of skin cancer has a high success in terms of starting treatment early, whereas late detection is far more serious.
The current process of detection either relies upon the subjective interoperation by medics, which is dependent upon the skill of the medic in ascertaining signs, followed up by laboratory testing should the medic feel that this is necessary.
To develop the artificial intelligence system, the researchers ‘trained’ the software using tens of thousands of skin images. Information relating to eumelanin (a chemical that gives skin its color) and hemoglobin (the red blood cell protein) levels was then fed into the system. The chemical data was drawn from biopsies. The aim here was to determine whether the machine could learn to detect cancer based on images alone, with the success tested by the corresponding chemical data. The device could then be used in medical practice to diagnose based on image alone without the need for lengthy and time consuming biopsies to be performed. A further downside with biopsies is that they require an invasive procedure, which requires specialist staff and is uncomfortable for the patient.
The image scanning has proved to be a success, with the machine providing a high level of accuracy. The technology will be commercialized and available to medics in 2018. The process, from design to testing to manufacture represents an example of how innovative start-up companies can emerge from the university sector.
Discussing the significance of the invention, lead researcher Professor Alexander Wong, from the University of Waterloo, said in communication with Digital Journal: “This could be a very powerful tool for skin cancer clinical decision support.”
The researcher added: “The more interpretable information there is, the better the decisions are.” This signaled the success and future development of machine learning to assist with medical diagnosis.
The research was presented at the 14th International Conference on Image Analysis and Recognition, which took place in Montreal during August 2017.
