Βy Prof. Daniele Regge, Chief of the Radiology Unit and Simone Mazzetti; Candiolo Cancer Institute, Torino-Italy
Presentation of the AI-QUAL study
Magnetic resonance imaging (MRI) has become an essential tool in the diagnosis and management of patients with localised prostate cancer (PCa). Present-day urology guidelines advise the use of MRI, as it can precisely and non‑invasively identify clinically significant prostate cancers that need treatment. This evidence has produced a rapid evolution and demand of prostate MRI worldwide, inevitably leading to variability in vendor and scan quality among imaging centres, with consequent risk of generating suboptimal diagnostic examinations. Adequate image quality is a prerequisite for the detection of prostate cancer, to guarantee an accurate visualization of the prostate gland and its surrounding structures, and it is related to both MRI technical parameters and to the patient’s preparation and habitus.
Efforts have been made to standardize image acquisition and interpretation via the development of internationally recognised scoring systems, such as the PI-RADS and PI-QUAL. However, both systems partially depend on the subjective experience and acumen of humans, which is related to readers experience and training.
Beside human interpretation and reporting of prostate MRI examinations, artificial intelligence (AI) is an evolving technology that is rapidly transforming the landscape of healthcare. AI’s role is not merely about automating processes; it fundamentally changes the approach to disease diagnosis, making it more precise and efficient, accelerating the analysis of medical images and reducing human interpretation errors. By learning from large datasets, AI algorithms can identify patterns and anomalies that might be overlooked by the human eye, which is crucial for diagnosing complex diseases such as cancer. However, the integration of AI in clinical settings is not without challenges. One concern is about training and validation of AI algorithms, based on adequate datasets satisfying the quality requirements set by the guidelines.
The AI-QUAL study we have designed within the ProCAncer-I project aims to develop an AI tool which could assist medical scientists with the selection of high-quality prostate MRI examinations through various steps. First, it will test if the MRI studies were acquired according to the minimum set of scanning parameters (e.g., slice thickness, and pixel resolution). Second, it will identify the most relevant artefacts or other issues that may be present in the images (e.g., patient movements, blurring, geometric distortions). These quality checks will ensure that MRI acquisitions are of sufficient quality for accurate AI development.
The AI-QUAL study was designed thanks to the collaboration between FPO, who provided the clinical expertise, Radboudumc who made available the platform and the interface to assess image quality, and FORTH who retrieved the dataset from the ProstateNET, balancing the cases according to MRI vendors and magnetic field strengths. Moreover, the AI-QUAL study will involve several other partners of the ProCAncer-I Consortium, including the 13 radiologists from 10 different Institutes who will assess the quality of about 1,000 prostate MRIs, setting the reference standard of the study by identifying possible artefacts or other aspects that could be detrimental for image interpretation. Then, the technical partners of the Consortium will implement the AI algorithms for the automatic identification of poor quality MRI examinations.
The results of the AI-QUAL study will allow to quickly evaluate MRI scans for adequacy for AI applications, without relying on personal opinions, allowing the creation of datasets compliant with the most recent urology guidelines, also increasing reliability of the algorithms that will be implemented and validated on such high quality MRI datasets.
