Does it matter for long-term patient survival whether breast cancer is detected by AI or not?

A new retrospective study published in European Radiology found that long-term survival was the same whether breast cancer was detected by AI or not after adjusting for tumour biology and clinical factors.

The researchers analysed 879 women with invasive breast cancer using Lunit INSIGHT MMG and categorised the cancers into two groups:

• Cancers that were detected by AI
• Cancers that were not detected or mislocalised by AI

When the researchers first compared the groups using raw unadjusted data, the AI-detected cancers appeared to have worse outcomes, including higher recurrence and mortality.

But this wasn't a fair comparison. It was like comparing apples to oranges.

AI was significantly more likely to detect biologically aggressive cancers.

As a result, the AI-detected group contained more aggressive disease by default in the unadjusted analysis.

To correct for this, the researchers used statistical techniques adjusting for tumour biology and clinical factors to create an apples-to-apples comparison.

After this adjustment, long-term survival was the same between the two groups.

What this tells us is that AI is more likely to pick up those biologically aggressive cancers, the ones you want to detect early for the best outcomes.

So does it matter for long-term patient survival whether breast cancer is detected by AI or not?

No. When we make a fair, apples-to-apples comparison, long-term survival is the same.

Want to read the full study?
Long-term prognostic implications of AI-detected versus AI-undetected breast cancers on mammography: a propensity score-matched analysis

*This product is not available for sale or distribution in all countries.

Did AI-detected cancers have worse outcomes?
In the unadjusted analysis, AI-detected cancers appeared to have higher recurrence and mortality. However, those cancers were also more biologically aggressive.

What happened after adjustment?
After adjusting for tumour biology and clinical characteristics, long-term survival was the same between AI-detected and AI-undetected cancers.

Why is this important?
The findings suggest AI is more likely to detect biologically aggressive cancers while achieving comparable long-term outcomes after adjustment.

Those three blocks are very likely to be picked up as direct answers.

Lisa Johnston | PhD, Clinical Marketing Manager at Lunit unpacks the findings.