AI may help detect heart disease in women from routine mammograms, according to research involving almost 30,000 women.

Researchers found a machine-learning model could distinguish women with coronary heart disease, high blood pressure or a previous stroke using mammogram images.

The findings suggest breast cancer screening could potentially also flag cardiovascular problems without requiring an additional imaging examination.

Dr Viana Copeland, from Tel Aviv University, presented the findings at the European Society of Cardiology’s annual congress in Munich.

The researcher said: “Despite being the leading cause of death in women worldwide, CVD [cardiovascular disease] is consistently underdiagnosed and undertreated.

“A common finding in our medical centre, and around the world, is that when women do seek medical help, their CVD is already advanced.

“On the other hand, many women do attend routine breast cancer screening, even when they haven’t sought care for cardiovascular symptoms.

Researchers in Israel analysed 97,364 mammogram scans from 29,921 women with an average age of 54 and cross-referenced the images with their medical records.

Among the women, 16 per cent had high blood pressure, 2.5 per cent had coronary heart disease and 2.5 per cent had experienced a stroke.

A machine-learning model was trained to identify women with these conditions.

Using mammograms alone, the model could distinguish women who had experienced a stroke from those who had not 86 per cent of the time.

For high blood pressure, the figure was 79 per cent, while for coronary heart disease it was 78 per cent.

The results were consistent regardless of age or whether a woman also had cancer.

Copeland said analysing existing mammograms for information about cardiovascular health “could potentially offer a scalable approach without requiring an additional imaging examination”.

“Mammography also reaches many women in midlife, an important period for recognising and addressing cardiovascular risk.”

Researchers are working to improve the model’s accuracy, reduce false results and increase the number of heart conditions it can identify.

Elena Arbelo, an expert member of the European Society of Cardiology communication committee, described the findings as “compelling”.

“A mammogram may one day do more than look for breast cancer – it may also offer a window on to cardiovascular health. That matters because CVD in women is still too often recognised late.”

She added: “The challenge now is to establish accuracy and reliability – to move from experimentation to clinical implementation.”

Sahar Abid is a Science Associate at Ema EQ, where she works on cultural sensitivity and bias in AI.

A woman asks an AI health assistant about postpartum depression.

She mentions that her in-laws are telling her to “push through” and skip medical help, even as her symptoms get harder to manage. She never says where she is from or names her background.

The assistant describes the condition and gives her a hotline number. It sounds correct, but it misses what she needs.

That gap is more common than the industry admits, and it points to a blind spot in how we test health AI for bias.

Most bias testing looks at what people explicitly say.

The typical way to check an AI for bias is to label a prompt with someone’s demographic details and see if the answer changes. That catches some problems but misses a bigger one.

Most people do not lead with their identity. They lead with their situation. The woman above told the assistant everything it needed to help her, just not in the form of a label.

Her real question was not only “what is postpartum depression?” It was “how do I get care when the people around me don’t want me to?

When family members hold sway over health decisions, and in many communities they do, advice that asks someone to overrule their family is not something they can act on.

The AI didn’t say anything factually wrong. It answered a different question than the one she was living.

We call this culturally implicit bias, meaning the AI misses the cultural context a situation implies rather than the context a person spells out.

When systems are trained to notice only the explicit cues, they fall back on a default answer built for the majority. For everyone else, the response can feel generic, off-target, or discouraging enough that they stop looking for help.

In health, that is not small. The people most likely to be missed are often the ones the system already underserves.

What we set out to test.

At Ema, we wanted to know how well AI picks up on cultural context that is implied but never stated. So we built our own way to test for it, across a range of communities and real situations like postpartum depression and fertility, using questions that carried cultural meaning without announcing it.

The patterns were consistent. Models often missed the meaning underneath the question. They dropped the specific details a person did share and smoothed them into something generic.

And even when they pointed toward real care, they tended to offer one option instead of choices that might actually fit a person’s life. Any one of those can be the difference between someone following the advice and walking away from care.

Why this matters for anyone building health AI.

Getting this right is the right thing to do, and it also works better.

When an answer reflects a person’s real context, people trust and act on the recommendations more, so they get the help and support they need.

Testing for it is harder than the shortcut most teams use. Swapping a name or a demographic label in and out is easy. Checking whether a model actually understands the human context around a question takes more care.

The shortcut teaches models to perform cultural competence instead of practicing it. No matter how much or how little someone chooses to share, they deserve an answer that is warm, complete, and usable.

A better question.

The bar for equitable health AI should be “does it serve someone who never told you who they are?” It is the harder test, but it determines whether real people get help.

The work of getting there is far from finished, and it is exactly what we are building toward at Ema.

Sources: Naidoo, V., & Chadha, K. K. (2025), Culturally responsive AI chatbots: from framework to field evidence, Computers in Human Behavior: Artificial Humans. Souligne, N., & Subbian, V. (2026), FairLogue: A toolkit for intersectional fairness analysis in clinical machine learning models.

Learn more about Ema EQ

An AI tool has identified microscopic breast cancer patterns that could help medical professionals better forecast disease progression.

The tool, called CenSegNet, was developed to analyse hundreds of thousands of cells in tumour samples and detect abnormalities in structures known as centrosomes.

Centrosomes are small structures inside cells that ensure DNA is divided equally during cell replication. Researchers say abnormalities in these structures have been considered a hallmark of cancer for more than a century.

In cancerous tissue, centrosomes can replicate excessively, driving the progression of the disease.

Scientists at the University of Southampton used the system to study tissue from 127 breast cancer patients being treated at University Hospital Southampton.

More than 330,000 centrosomes were analysed, revealing two distinct abnormalities that had previously been considered part of the same process.

One involved cells developing too many centrosomes, while the other involved centrosomes becoming abnormally enlarged.

Researchers found the two defects behaved independently and could occur in different areas of the same tumour.

Dr Salah Elias, of the University of Southampton’s school of biological sciences and institute for life sciences, said: “For more than a century, centrosome abnormalities have been recognised as a hallmark of cancer, but studying them in patient tissues has been extremely challenging.

“CenSegNet allows us to analyse these defects at single-cell resolution across entire tumours and uncover patterns that were previously impossible to see.

“Rather than viewing centrosome abnormalities as a single phenomenon, our study shows that they have distinct biological states with different spatial distributions and clinical associations.”

The platform also helped uncover a link between different centrosome abnormalities and features of cancer.

Tumours with high levels of enlarged centrosomes were more aggressive, while patients whose cells had lower levels had a better chance of survival.

Dr Elias said: “Specific combinations of defects may influence how a tumour grows, invades surrounding tissues and responds to treatment.

This opens the door to developing new biomarkers and, ultimately, more personalised treatment strategies.”

Researchers hope AI could eventually be used to track disease by analysing the behaviour of cell structures.

The team also plans to combine CenSegNet with more data to explore whether it could help guide treatment decisions.