An interview with Prof. Philippe Puech

As the European Congress of Radiology (ECR) 2026 approaches, prostate cancer imaging is entering a new phase. Multiparametric prostate MRI, structured reporting, and AI in radiology are redefining how clinicians detect and manage disease, with a growing focus on reducing unnecessary biopsies and improving PI-RADS consistency.

Ahead of Europe’s leading radiology congress, we spoke with Prof. Philippe Puech, Professor of Radiology, an international authority in prostate MRI, and advisor at Quibim, about how imaging and AI are shaping prostate cancer diagnosis.

 

Fewer biopsies through smarter prostate MRI

Do you think we are still performing too many biopsies on low-risk patients, even in an era where MRI is prioritized? How can imaging help avoid unnecessary procedures?

Yes and no. For over 20 years, all patients at Lille University Hospital have undergone multiparametric MRI before biopsy. In the past, many patients with a negative MRI were still biopsied because of elevated PSA alone.

Over the last five years, practice has changed. Risk calculators such as ERSPC and updated EAU guidelines now allow us to defer biopsy in selected cases. Patients with PI-RADS 1–2 often avoid biopsy, PI-RADS 3 cases are discussed in context, and PI-RADS 4–5 receive targeted biopsies.

As a result, more than 40% of patients who would previously have been biopsied can now avoid or defer the procedure.

However, two safeguards are critical: robust MRI interpretation — via expert reading, double reading, or AI assistance — and the use of MRI-ultrasound fusion for targeted biopsies rather than cognitive guidance alone.

Why structured reporting matters in prostate cancer care

How important is structured reporting today in reducing variability between centers and radiologists?

Structured reporting is fundamental in prostate MRI. A report must go beyond lesion detection and standardize how size, aggressiveness, extracapsular extension, and image quality are described using systems such as PI-RADS, PI-QUAL, PRECISE, and PI-RR.

In our software, nearly 50 variables are documented and organized in a logical framework so radiologists and urologists can quickly retrieve precise information. This approach significantly reduces inter-observer variability.

Including the standardized PI-RADS sector map is indispensable, as it minimizes localization errors and optimizes targeted biopsies.

 

AI for consistent PI-RADS interpretation

At conferences such as ECR, we see the true potential of technology in the medical field. What role can AI really play in improving consistency in PI-RADS interpretation and report quality?

AI can assist at several stages of prostate MRI. It can automatically assess image quality, detect motion artifacts, verify acquisition parameters, and compute AI-driven PI-QUAL scores.

During interpretation, AI can pre-detect clinically significant lesions, suggest PI-RADS scores, evaluate aggressiveness, and help resolve equivocal PI-RADS 3 findings.

For reporting, AI can select key images, generate standardized sector maps, compare exams in active surveillance, and check reports for missing measurements.

Many of these technologies already exist; however, they are not yet harmonized, properly calibrated, externally validated, uniformly integrated into commercial platforms, or widely democratized across clinical practice. Considerable work remains to be done in these areas to be widely adopted.

 

Barriers to routine AI adoption

From your clinical perspective, what are the main barriers that still prevent AI tools from being routinely used in prostate cancer diagnosis, and which of them do you consider the hardest to overcome in daily practice?

One major barrier is inconsistent usage. Even when AI is available, radiologists apply it differently. A clinical AI usage charter is therefore needed to define how and when AI should support interpretation.

While most AI solutions show strong performance, they are trained to detect and quantify different endpoints and often function as “black boxes.” This lack of calibration and transparency makes it essential to define adjustable thresholds and establish clear, interpretable performance metrics.

AI must also integrate seamlessly into PACS and RIS without slowing workflows. Finally, an under-discussed issue is the financial and energy costs of deploying multiple AI tools, which inevitably limits their widespread availability due to resource constraints.

 

Aligning radiology and urology with AI

How do you think AI platforms such as QP-Prostate can help standardize prostate MRI evaluation and better harmonize radiology with urology in current clinical practice?

Urologists greatly value standardized prostate MRI reports. When the clinical request is clear and the radiologist provides a structured report with lesion descriptions, PI-RADS sector maps, and annotated images, collaboration improves.

QP-Prostate supports this by delivering automated quality assessment with an AI-QUAL score, precise prostate volume and PSA density calculations, lesion assistance, and automated reporting directly in PACS.

It also empowers less experienced radiologists to identify relevant lesions while increasing confidence in borderline cases—either confirming true positives or preventing unnecessary overcalls. This consistency reduces inter-observer variability. Beyond interpretation, the platform elevates patient care by adding a safety layer through automated second reading.

 

Advice for young radiologists

Finally, what message would you give to young radiologists who want to specialize in prostate imaging in an AI-driven era?

Do not fear AI — master it. Understand its strengths, limitations, and performance metrics. Learn when the human reader outperforms the algorithm and how to manage discordance between your interpretation and the software output.

The radiologist remains a fully-fledged clinician. In many cases—and I see this every day—a careful consideration of the patient’s overall clinical history, comorbidities, and full medical record often proves decisive, going beyond image analysis alone.

Artificial intelligence is a powerful form of assistance that it would be unwise to forgo, but it must never be allowed to take full control. It helps reclaim time for deeper interpretation and patient communication, but the human physician must always retain critical judgment. As Rabelais famously wrote in 1532: “Science without conscience is but the ruin of the soul.” We should keep that in mind.

 

 

At ECR 2026, Quibim will showcase how AI-powered prostate MRI can reduce unnecessary biopsies, standardize PI-RADS reporting, and improve prostate cancer diagnosis.

Join us to explore how technology and clinical expertise are shaping the future of prostate imaging.