Digital Pathology and AI: How Cancer Slides Are Becoming Data

From glass slide to image file

For more than a century, a pathologist looked at tissue through a microscope. Now that glass slide can be scanned into a very large image file, and software can examine it.

FDA describes digital pathology as making digital scans of tissue from slides. The data can then be viewed, analyzed, sent, stored, and read. The change is not cosmetic. Once a slide is a file, it can be shared with a distant expert. It can be kept, looked at again, and measured by a computer.

This page explains federal sources. It is not medical advice and does not suggest a test or treatment.

What the pathology report still does

Start with what the software is being added to.

NCI explains that a pathology report describes a tissue sample taken from a patient. It is written by a pathologist. That is a doctor trained to identify disease by studying cells and tissues under a microscope. The report usually has three parts. A description of what the sample looks like. A description of what it looks like under the microscope. And a final diagnosis.

Then NCI states the point that matters most here. The pathology report provides the definitive cancer diagnosis. It is also used for staging and to help plan treatment.

Getting there often takes more than the first slide. Extra stains such as immunohistochemistry may be needed. So may molecular tests, expert review, or a fresh sample. Our page on understanding your pathology report walks through the sections.

What has actually been authorized

Two milestones are worth knowing, because coverage tends to blur them.

FDA allowed marketing of the first whole slide imaging system. It lets pathologists review and interpret digital slides made from biopsied tissue. An FDA official said it lets pathologists read tissue slides on a screen to make diagnoses, rather than under a light microscope. That decision was about the imaging system. It was not about AI.

Separately, NCI notes that FDA has authorized marketing of AI-based software that helps pathologists identify areas of prostate biopsy images that may contain cancer. Read that carefully. The software helps a pathologist find areas to look at. It does not issue the diagnosis.

Where these systems tend to break

A tool can do beautifully on clean research images and still stumble in a working lab. Staining varies. Slide thickness varies. Scanners differ. Tissue gets folded, torn, or poorly fixed.

So a strong test compares the whole workflow with the current process. It should include hard cases and poor-quality slides on purpose. It should record where humans and software disagreed. It should check the final diagnoses. Testing at outside labs shows whether results carry over. Monitoring should go on after any scanner or software change.

The intended task also has to be stated. Detection, measurement, and prediction are different jobs with different evidence behind them.

The counterweight: false flags and finding more than matters

Detection stories rarely include this half, and they should.

A more sensitive tool flags more areas. Some of those will turn out to be nothing. NCI describes what follows a false-positive result. Anxiety, then more tests and procedures that carry risks of their own. In a pathology workflow that can mean extra stains, deeper cuts, or another biopsy. It can also mean a report that sounds worse than the tissue warrants.

There is a quieter problem too. NCI describes overdiagnosis as finding a cancer that would never have caused symptoms or harm. Overtreatment is the treatment that follows it. A tool tuned to catch every small change can raise both. NCI's blunt line about screening fits here as well. Finding the cancer may not improve the person's health.

So "the AI found more" is not automatically a better result. The question is whether the people involved ended up better off.

What stays with the pathologist

• An AI mark on a slide does not diagnose cancer by itself.
• A system trained on one laboratory's slides may not behave the same elsewhere.
• Guesses about risk or treatment response need their own testing in real care.
• The final diagnosis, and the responsibility for it, remain with the pathologist and the care team. How cancer is diagnosed explains the wider process.

Questions about an AI pathology headline

• Was the software used for detection, measurement, or prediction?
• Did pathologists review the output, and did the diagnosis change?
• Was it tested across different laboratories, scanners, and patient groups?
• How many flagged areas turned out to be nothing?
• What happens to monitoring after a software update?

How this article was prepared

An AI-assisted editorial system helped prepare this page. No named medical reviewer has reviewed it unless one is listed.

The National Cancer Information Foundation publishes Cancer Explained. This page is for learning. It is not medical advice and does not suggest a test or treatment.

See an error, old source, or unclear wording? Tell us.