AI shows promise in ovarian cancer detection within NHS

Recent research from Sweden has indicated that artificial intelligence (AI) could play a pivotal role in detecting ovarian cancer within the National Health Service (NHS). The study suggests that AI technology is capable of identifying small tumours, or lesions, on ultrasound images with a success rate of nearly 90%, outperforming ovarian cancer specialists who achieve an 80% accuracy rate.

Ovarian cancer is widely recognised as one of the more challenging malignancies to diagnose. Its symptoms—such as bloating, frequent urination, vaginal discharge, and constipation—are often confused with less serious medical issues. Compounding the problem, there is currently no effective screening method for the disease, which often leads to diagnoses occurring only after the cancer has advanced, with research indicating that approximately 80% of cases are identified only after metastasis.

In the UK, around 7,500 women are diagnosed with ovarian cancer each year, with approximately 4,000 fatalities annually. The potential adoption of AI for this purpose could significantly alter the landscape of diagnosis and treatment in the NHS, particularly in enhancing the speed and accuracy of cancer detection.

The NHS’s previous initiatives have already explored AI technology, including a breast screening trial initiated last year aimed at enhancing mammogram accuracy through AI analysis. Researchers at the Stockholm South General Hospital executed the latest study by uploading over 17,000 ultrasound images of ovaries to a self-learning AI programme, commonly referred to as a neural network model. These images encompassed both cancerous lesions and benign growths.

The AI demonstrated a remarkable capacity to differentiate between the two, establishing a groundbreaking benchmark in the diagnosis of ovarian cancer. The findings of the study suggest that incorporating AI into hospital practices could enhance the daily capacity of referrals managed by doctors by approximately 60% and could potentially reduce misdiagnoses by nearly 20%.

Professor Elisabeth Epstein, a senior obstetrics and gynaecology consultant at Stockholm South General Hospital, noted, “Ovarian tumours are common and are often detected by chance. This suggests that neural network models can offer valuable support in the diagnosis of ovarian cancer, especially in difficult-to-diagnose cases and in settings where there’s a shortage of ultrasound experts.”

As the healthcare industry contemplates the subsequent integration of advanced AI technologies into clinical practice, this research reinforces the promising prospects for AI’s contributions to oncological diagnostics and patient care outcomes.

Source: Noah Wire Services