AI Model Detects Prostate Cancer with Radiologist-Level Accuracy
An AI model developed for detecting prostate cancer has demonstrated performance comparable to that of experienced radiologists. A team, including researchers at the Mayo Clinic in Minnesota, US, found that the model could serve as a "potential assistant" to radiologists, helping to improve diagnosis from magnetic resonance imaging (MRI) scans by increasing detection rates and reducing false positives.
Key Points:
- Multiparametric MRI: Radiologists typically use multiparametric MRI for diagnosing prostate cancer, which provides a more detailed picture of the prostate gland than standard MRIs.
- PI-RADS Score: Results are expressed as a PI-RADS score (Prostate Imaging-Reporting and Data System), with a higher score indicating a higher likelihood of clinically significant cancer.
- Study Findings: The study, led by Naoki Takahashi from the Department of Radiology at the Mayo Clinic, highlighted the difficulty in interpreting prostate MRI and the higher diagnostic performance of more experienced radiologists.
- AI Training: The researchers trained a convolutional neural network (CNN) to predict clinically significant prostate cancer from multiparametric MRI. The model's performance was compared with that of abdominal radiologists in patients who had undergone MRI but without known clinically significant prostate cancer.
Model Performance:
- In a retrospective study involving 5,215 patients (5,735 examinations) who underwent multiparametric MRI for prostate cancer evaluation, the AI model's performance in detecting clinically significant prostate cancer was found to be on par with experienced radiologists.
- The model could be used as an adjunct to radiologists, enhancing the accuracy of prostate cancer detection.
Global Impact:
A Lancet Commission on prostate cancer projected that global cases could more than double between 2020 and 2040, with deaths increasing by 85%. Low- and middle-income countries are expected to bear the brunt of this spike. The commission advocated for evidence-based interventions, such as early detection and diagnosis, to save lives and prevent ill health from prostate cancer in the coming years.
Dr. Takahashi emphasized that while the AI model should not be used as a standalone diagnostic tool, its predictions could significantly aid radiologists in the decision-making process, potentially improving clinical outcomes and patient care.