NHS could replace doctors with AI to detect ovarian cancer after study shows software is more effective at detecting hard-to-find diseases

Artificial intelligence could be used by the NHS to detect ovarian cancer after new research found the software was more effective than doctors at detecting the elusive deadly disease.

The technology, also known as AI, can correctly detect small tumors, called lesions, on ultrasound images of the ovaries almost nine times out of 10, according to Swedish researchers.

In contrast, ovarian cancer specialists discovered the lesions in only eight out of ten cases.

Experts say the findings are significant because ovarian cancer is considered one of the most difficult types of tumors to diagnose.

This is because the symptoms – which include bloating, frequent urination, vaginal discharge and constipation – are often mistaken for signs of less serious illnesses.

There is also currently no effective way to screen women for the disease. This means that by the time the cancer is noticed, it has often spread throughout the body.

Research shows that as many as four in five cases are only discovered after the cancer has spread to other parts of the body.

Around 7,500 women in Britain are diagnosed with ovarian cancer every year – and around 4,000 die from it in the same period.

The technology, also called AI, can correctly detect small tumors on ultrasound images of the ovaries almost nine times out of ten (Stock Image)

Research has already shown that artificial intelligence can speed up the diagnosis of patients with skin and lung cancer.

Last year the NHS announced a first-of-its-kind breast screening trial, using AI to look for signs of cancer on mammograms, in a bid to improve the accuracy and speed of diagnosis.

The latest study, published by scientists at Stockholm South General Hospital, uploaded more than 17,000 ultrasound images of ovaries to a self-learning AI computer program – also known as a neural network model.

These images showed some patients with cancerous lesions and others with growths that were not cancerous – also called benign lesions.

After analyzing all images, the AI ​​was able to correctly identify the signs of ovarian cancer in the vast majority of cases.

The researchers concluded that the use of AI in hospitals, because of its speed and accuracy, could speed up the number of referrals doctors see every day by around 60 percent and reduce the number of misdiagnoses by almost a fifth.

“Ovarian tumors are common and are often discovered incidentally,” says Professor Elisabeth Epstein, senior consultant in obstetrics and gynecology at Stockholm South General Hospital.

‘This suggests that neural network models can provide valuable support in the diagnosis of ovarian cancer, especially in difficult-to-diagnose cases and in settings where there is a shortage of ultrasound experts.’