India-First-Global-Insights-Analysis -Sharing-PlatformIndia-First-Global-Insights-Analysis -Sharing-Platform

Machine learning to help doctors predict the success of epilepsy brain surgery

, August 4, 2013, 0 Comments

Researchers in Germany have developed a technique, which allows them to predict the chances of success for a surgical procedure to treat epilepsy. It could prevent unnecessary surgery.

Epilepsy is a common neurological disorder. It occurs when many nerve cells fire simultaneously in the brain – leading to seizures. It affects around 50 million people in the world.

There are drugs that can help manage epilepsy, but some patients experience resistance to the treatment. For them, the only hope is surgery – the affected temporal lobe in the brain is often removed.

But the surgery doesn’t always work – about a third of patients experience little or no improvement after the procedure. They face the threat of seizures for the rest of their lives.

“It affects their quality of life, it can affect their employment and their education opportunities and it can play a big part in their own life and that of their family,” Louise Cousins from UK-based Epilepsy Action told DW.

Undergoing the surgical procedure presents a daunting decision for many patients. Apart from the risks associated with all surgery, epilepsy patients also have to deal with the fact that there’s a 30 percent chance of the procedure failing.

Predicting surgery success

But this may soon change. Scientists from the Bonn University Hospital and the Max Planck Institute for neurological research in Cologne have come up with a technique which has enabled them to predict the success of the procedure for temporal lobe epilepsy in 9 out of 10 cases.

The system is based on machine learning algorithms, which allow computers to learn from data.

Using a computer program developed by Max Planck mathematician Delia-Lisa Feis, epilepsy specialists compared the MRI images of patients who improved following the surgical procedure and those of patients whose operations failed.

The idea was to develop a classifier – to take structural information and “train a classifier to have a diagnostic capability,” Feis says. “We trained an algorithm to see the differences.”

More tests needed

Despite the promising results so far, the researchers say it will be some time before the technique is used on patients.

“We are still at a stage that is too early for clinical application – further research is needed,” said Professor Christian Elger, director of the Bonn Epilepsy Clinic, in a press release.

Professor Hajo Hamer from the Epilepsy Center in Erlangen, Bavaria, who didn’t participate in the research, agrees.

“It’s a good idea, but it now has to be proven with a prospective study,” Hamer told DW.

The tests were performed on patients who had already had the surgery. The scientists will now study how the system works on patients who have yet to undergo the procedure.

Professor Hamer says there maybe more than one solution. “You may need to combine several methods to get a higher accuracy,” he said.

Machine-learning in medical diagnosis

While Professor Elger and his colleagues carry out further tests, mathematician Delia-Lisa Feis is developing a similar system to help neuroscientists in the treatment of Parkinson’s disease.

With the arrival of big data – collections of large data sets – in hospitals, machine learning is likely to find its way into medical diagnostics.

By looking at the similarities and differences between a large sample of patients, doctors could potentially have a better sense of how a given treatment will affect a patient. And that’s something that will be welcomed by both medical practitioners and patients – the patients won’t have to undergo unnecessary treatments that do not lead to a cure.

Source: Deutsche Welle | www.dw.de






About author
Deutsche Welle (DW) is Germany’s international broadcaster, Headquarters in Bonn and Berlin, having full range of presence in television, radio and online services. DW is known for its in-depth, reliable news and information in more than 30 languages ...more