Findings that support the greater accuracy of computerised prognoses for bladder cancer compared to those made by doctors, has been presented to the American Association for Cancer Research by Professor Raouf Naguib of Coventry University.

Addressing the 90th Annual Meeting of the Association in Philadelphia, Raouf Naguib, Professor of Biomedical Computing, outlined research that he and his team have conducted into the use of computer software incorporating neural networks, to produce prognoses for patients with bladder cancer.

The capacity of artificial neural networks to recognise complex patterns within data and to analyse large numbers of variables has already been used in to provide prognoses for conditions such as back pain, infertility and prostate cancer. However, this is the first time that this technology has been applied to bladder cancer.

The disease is the fourth most common cancer among American men, with a similar incidence among men in Britain. Until now, accurate methods of predicting how the disease will progress, the likelihood that tumours will recur, and the survival rate of patients with the disease have all been lacking.

Coventry University researchers used data from 210 patients with newly-diagnosed bladder cancer, and trained the artificial neural networks to analyse it for 15 markers, including gender, age, tumour stage and grade, and epidermal growth factor.

They then asked four experienced consultant urologists to make prognoses on the same data, and compared their predictions with those of the neural network. The results showed that the more accurate prognoses were produced using neural networks.

For patients with superficial cancer, the prognosis of the progress of the disease produced by the neural networks was 80 per cent accurate as opposed to 74 per cent for the urologists.

When asked if a patient with advanced cancer would be alive in a year’s time, the neural networks were 82 per cent accurate, compared to only 65 per cent for the urologists. In those patients the five-year survival rate is usually around 50 per cent, so the level of accuracy achieved by the neural networks could make a huge difference to those individuals affected.

Professor Naguib believes that neural networks will become an important tool for doctors in years to come, but warns of their limitations.

"The decision-taking of the neural networks takes just seconds", he explains, "but training it takes weeks or months and requires a large amount of data."

He admits the findings are promising but states: