Using artificial intelligence to measure image quality in Magnetic Resonance Imaging
What you need to know
- Magnetic Resonance Imaging (MRI) scans are an important step in diagnosing prostate cancer and can help avoid an invasive biopsy for men who have no evidence of cancer on their scan. However, the quality of MRI images can vary significantly, which can make interpretation extremely difficult.
- We’re funding Dr Barrett to develop and test a software tool which uses artificial intelligence (AI) methods to measure the quality of MRI images.
- His research will help improve the quality of MRI scans across hospitals, which will allow for more accurate diagnosis of prostate cancer.
MRI is a key step towards diagnosing prostate cancer
MRI scans can be used to find out if prostate cancer is present before a biopsy. If no area of concern is seen on the scan, a biopsy may not be needed. If it appears that cancer is present, the MRI scan can help doctors target the biopsy needle to areas of concern.
However, the quality of MRI scans can vary significantly between centres across the country, and a poor-quality image can make it harder for doctors to tell if cancer is present. So, a good quality image is important in figuring out whether a man has prostate cancer or not.
Improving the quality of MRI scans with AI
In this project, Dr Barrett and his team will develop and test a software tool which uses AI to assess how good an MRI image is. They will be helped by Lucida Medical, an AI software company that specialises in the analysis of MRI for prostate cancer.
They aim to develop a scoring system by working with prostate cancer MRI experts and looking at MRI scans that have already been taken. They’ll set up a 1-5 scoring scale, so if an image scores a 5 this would mean it is of good quality, with a score of 3 being acceptable. The next stage will be to teach the AI tool how to score the quality of the images based on this scale and test it on previously taken MRI scans.
Once they have confirmed that it is working, they will test the AI tool in the clinical setting. They will see if this tool can be used to tell doctors if there is a problem with the MRI procedure when the image is of poor quality, and whether a patient needs to be rescanned.
Supporting better diagnosis
By the end of this project, the software should provide a benchmark for good quality images and help drive improvements in diagnosis. This system should ensure that the MRI scans are consistently performed to a high quality, wherever the scan is done.
This tool may in the future be able to provide real time feedback during the scan and so reduce the need for men to go back to the hospital for repeat scans. Ultimately, this tool should improve the use of MRI across centres which will greatly support the diagnosis of prostate cancer.
Reference – RIA19-ST2-014
Researcher – Dr Tristan Barrett
Institution – University of Cambridge
Award - £300,934