AI algorithms are able to recognise early signs of prostate cancer in more than four out of five (80%) cases where pathologists previously missed them. Using AI, researchers at Uppsala University have been able to detect subtle tissue changes that allow the cancer to be detected long before it becomes visible to the naked eye.
In collaboration with Umeå University, researchers analysed biopsies from 232 men who had previously been assessed as cancer-free. Within 2.5 years, half of them developed aggressive prostate cancer. Using artificial intelligence, subtle tissue changes that had not been recognised beforehand were subsequently identified in the men who later became ill. ‘Our AI was able to recognise signs of cancer that pathologists did not see in more than 80 per cent of cases,’ said Professor Carolina Wählby, who led the AI research published in Scientific Reports.
Previously missed diagnoses
The researchers developed an innovative training method for the AI, as all tissue samples were originally classified as “negative”. By analysing images of tissue samples from men who later developed cancer, the system learned to distinguish abnormal patterns that indicate early malignancy.
The findings suggest that AI analysis of standard biopsies can be a valuable addition to existing screening methods. In this way, AI can contribute to personalised follow-up of men who are considered healthy based on standard diagnostics.
The data and algorithms used have been made available for further scientific development. This paves the way for future AI applications in the early diagnosis of prostate cancer.
Faster prostate cancer diagnosis with AI
Last year, an international study coordinated by Radboudumc already showed that AI can detect prostate cancer more accurately and faster than radiologists. By using AI, the number of unnecessary biopsies is significantly reduced, as the system indicates an incorrect suspicion half (50%) as often. The study, called PI-CAI, involved more than 200 AI teams and 62 radiologists from 20 countries, who analysed over 10,000 MRI scans. The five best-performing AI models were combined into a powerful “super algorithm”.
When compared to radiologists, AI detected 7 per cent more clinically relevant prostate cancers. Due to the growing demand for MRI diagnostics and the shortage of radiologists, AI can play an important role in alleviating the workload. Although the results are promising, the AI still needs to be further validated and is not yet ready for use in clinical practice. Follow-up research is needed to confirm its applicability in everyday healthcare.
AI-driven treatment
AI has also already proven its added value in the treatment of prostate cancer. Earlier this year, Isala Hospital took an important step towards personalised radiotherapy for prostate cancer with the commissioning of the Elekta Evo CT linear accelerator. This advanced technology makes it possible to visualise tumours with razor-sharp precision and to adapt the treatment plan to the patient's current anatomy during radiation.
Thanks to AI and improved cone beam CT, doctors can better distinguish between the tumour and surrounding tissue, leading to more precise and effective treatments with less damage to healthy tissue. The Evo will initially be used for tumours in the pelvic area, such as prostate cancer. Isala will replace all accelerators in the coming years to apply this approach on a broad scale.
