DOGS – Computerised image analysis for rapid and exact diagnosis of prostate cancer

AI improving diagnosis of prostate cancer

Prostate cancer is the most common form of cancer in men. In Sweden, around 10 000 men are struck down each year. When a pathologist makes a diagnosis using a microscope, there is also an assessment of the cancer cells’ growth pattern – these differ, depending on how aggressive the cancer is. This estimate, which is not easy, is of great importance for giving the patient the right treatment and follow up, comments Anders Bjartell, consultant and professor at Lund University.

Uncertain diagnosis

“There is a shortage of experienced specialist doctors and the number of patient samples has increased significantly in recent years, often 12 samples or more are taken per patient. And naturally there are subjective judgements – studies have shown that up to one in five cases do not get the same assessment result when made by different doctors”, says Anders Bjartell.

DOGS (Digital pathology for Optimized Gleason Score in prostate cancer) is a research project in which researchers have developed an image analysis program that uses artificial intelligence.

Trained software

The program has been trained to recognise prostate cancer cells and their growth patterns (tumour grade according to Gleason) by scanning in a large number of tissue samples that have been assessed by pathologists using regular microscopy. The next step in the project is to validate how reliably the programme has learned to recognise cancer cells and their growth patterns in new patient samples.

“At present, it’s not possible to replace the pathologist, but interactive decision-making support can be of great assistance for faster and more reliable diagnosis”, says Anders Bjartell.

Large market

Marketing of the program to healthcare is to be managed by the medtech company, Sectra, a spin off from research at Linköping University and a pioneer in image processing for healthcare. Sectra deems that healthcare in general is quite ready to manage decision-making support such as DOGS, as it is becoming increasingly common for hospitals to scan in images from microscopy in computer systems.

“It would be easy for the doctors to start using our program, perhaps within a couple of years. The market for decision-making support in digital pathology could be very large, especially if you reckon with other image analyses that the program can also be used for”, states Anders Bjartell.

Text updated 25 July 2017

Text: Elisabet Ottosson

Kontakt

Projektledare