Harnessing the power of computers to improve prostate cancer detection

The Department of Radiation Oncology/Medical Physics at Calvary Mater Newcastle has developed a machine learning approach for the identification of prostate cancer from magnetic resonance imaging (MRI).

This newly developed artificial intelligence system is able to identify more than 100 features within the MRI images that can differentiate cancer from healthy prostate tissue. Many of these features are invisible to the human eye and consequently before now would not be used in the decision making process. However, by putting the computer alongside the human expert (in this case a radiologist or radiation oncologist) this additional information facilitates improved decision making and ultimately provides a better patient outcome through more targeted treatment.

John Simpson, Chief Medical Physicist, Calvary Mater Newcastle, said,

“Radiation oncology increasingly uses medical images to identify and target cancer for more accurate treatment. However, making the most of this imaging information is not straight forward and this is where computers can help. Not only can a computer process a lot of information but unlike a human being, the computer is not limited by the human eye and can extract and analyse information within an image that is invisible to you and I.”

John explained, “This research involved showing the computer a number of examples of MRI scans with the prostate cancer identified and then asked the computer to learn how to recognise cancer from healthy tissue. The ultimate aim of this work is not only to help improve cancer detection but to identify cancer characteristics associated with treatment outcomes. Knowing this type of information it is hoped will lead to a personalised treatment approach where each patient receives a prescription that is tailored to their individual characteristics rather than the more general approach that is today’s standard of care.”

The team is now aiming to improve this system with the use of additional types of scans and then plan to apply this learning to other types of cancers.

This research was made possible due to the generosity and memory of Rodney Sorenson and the Tomaree Prostate Cancer Support Group whose valuable support allowed the Department of Radiation Oncology to undertake this pilot study.