Radiologists or AI? A Gorilla Too Big to Ignore

We Often Fail to See What We Do Not Expect

In 2013, a striking experiment shook the medical community. Computed tomography (CT) images were shown to 24 radiologists trained to detect lung nodules. Some images contained a digital gorilla deliberately inserted by researchers, 48 times larger than a typical nodule. The result was surprising: 83% of the radiologists missed the conspicuous gorilla. Eye tracking confirmed that they had looked directly at its location. Why did they fail to notice it?

This illustrated “inattentional blindness.” The human brain can filter out unusual or unexpected information, particularly while concentrating on a specific task. The experiment demonstrated that even highly trained specialists may miss visual clues under information overload.

Today, a similar question has emerged: Can AI help us notice the “gorillas” right in front of us?

The Rise of AI in Imaging

Artificial intelligence (AI) has achieved notable success in tumor detection, but its role in radiology is much broader. Current algorithms can:

• Instantly analyze CT images for signs of stroke.

• Flag fractures in trauma imaging.

• Quantify even small changes in tumor size on oncological imaging.

Particularly in oncology, AI can process large imaging datasets, identify details radiologists may overlook, and support clinical decisions.

MASAI and EDITH UK: AI Enters Clinical Trials

Important clinical studies are assessing AI's potential in the real world:

• The MASAI Study: The Lancet Digital Healthpublished this randomized controlled study investigating whether AI-supported mammography screening can improve early breast cancer diagnosis. The aim is to reduce false negatives and interval cancers.

• The EDITH UK Project: This large study based in Englandtests whether AI can perform as well as, or better than, humans in breast cancer screening. If successful, it could shorten diagnosis and reduce waiting times.

These examples show that AI has moved beyond experimental technology and is becoming a cornerstone of modern diagnosis .

Risks Alongside Benefits: Data Privacy and Algorithmic Bias

Like any powerful technology, AI carries risks requiring careful management:

• Data Bias: Algorithms trained only on data from large urban hospitals may not be sufficiently accurate for rural or disadvantaged populations.

• Data Security: As cloud-based systems become more common, patient confidentiality becomes even more important. Strong encryption, restricted access, and regulatory compliance are essential.

• Automation Bias: Blind trust in algorithmic outputs may lead physicians to incorrect diagnostic and treatment decisions.

AI should therefore be used carefully, with final decisions always left to human expertise.

The Human Touch: Moravec's Paradox and Cognitive Agility

Moravec's paradox highlights that AI can readily perform complex calculations while still struggling with skills that seem easy to humans, such as empathy, intuition, and interpreting context. For example:

• AI can detect a nodule,

• But cannot assess the patient's life story, emotional state, and social environment.

This is where cognitive agility matters: Clinical decision-makers need both technical analytical skills and empathy.

Radiologists and AI: Collaboration Rather Than Competition

Although the title asks “Radiologists or AI?”, the future lies in the formula “Radiologists + AI.” In this collaboration:

• AI rapidly analyzes images,

• Flags preliminary findings,

• And the radiologist interprets them within the patient's clinical context.

This hybrid approach improves diagnostic accuracy while reducing radiologists' mental workload. Studies such as MASAI and EDITH show how earlier AI-assisted breast cancer detection can translate this collaboration into patient benefit.

The Future Radiologist: AI-Literate and Adaptable

As AI takes on more tasks, radiologists inevitably need new skills. The future radiologist requires 4 core competencies:

1. Obtain AI Training: Gain hands-on experience through workshops, online certification programs, and in-house practical training.

2. Promote AI Literacy: Advocate integrating ethics, data science, and AI evaluation into specialist training.

3. Demand Equity and Safety: Work with scientists and regulators to reduce algorithmic bias and protect patient data.

4. Stay Alert: Treat AI output as the starting point of a decision, remembering that the physician remains the final decision-maker.

Conclusion: Let Us Stop Missing the “Gorillas” in Imaging

From the famous 2013 gorilla experiment to pioneering studies such as MASAI and EDITH, radiology's way of seeing has changed. We now have both advanced algorithms and experienced radiologists.

When implemented appropriately, AI can:

• Improve clinical accuracy,

• Shorten diagnosis,

• Help patients receive test results sooner.

How we use this technology will determine how effective it becomes.

What Do You Think?

Is AI truly revolutionizing radiology? Or, without sufficient care, will we continue missing new “gorillas”?

Share your thoughts so we can see more together through this transformation.