New AI method improves early breast cancer detection

AI is already proving useful for detecting cancers with high accuracy — improving patient outcomes and saving lives.

The earlier doctors detect breast cancer, the higher the chances of beating it.

AI algorithms are so good at detecting breast cancer from radiology scans that they're even starting to outperform humans in some cases — the UK's National Health Service (NHS) is now using AI to analyze mammograms, helping to spot breast cancer cases missed by human doctors.

But while AI can speed up the analysis of cancer screenings, doctors also need new methods to identify the markers of cancer at the earliest stages, before cancer tumors are visible in scans.

A pilot study has trialed a new method of blood screening, combined with AI, to detect breast cancer at very early stages.

The study, which was published in the Journal of Biophotonics in November, found their method could detect stage 1A breast cancers with an accuracy of 90-100%.

"This study is a milestone in being able to identify subtypes of breast cancers at very early stages, with high accuracy. Early cancer diagnosis saves lives — that's why our study is important," said study lead author Kevin Saruni Tipatet from the University of Edinburgh, UK.

However, Tipatet said the approach was only tested on 24 patients and isn't ready to be used in a hospital setting until they confirm the results in larger-scale studies.

New approach for detecting breast cancer early

Conventional methods of cancer detection are based on detecting markers of a cancerous tumor itself.

X-ray mammograms might be used to identify changes in breast tissue that indicate cancer, often before cancer symptoms appear. Or biopsy tests may detect molecular signatures of cancer cells in the body.

But these methods can often miss early-stage cancers — it's "like looking for a needle in a haystack", Tipatet told DW.

"Most technologies are focused on finding that needle, but they don't look at the whole picture around it."

Tipatet's study trialed an approach that gathers information about how the body responds to the cancer.

Rather than just looking for cancer tissue, they searched for "molecular fingerprints” that indicate the body is fighting breast cancer.

These fingerprints are "from the cancer itself, or the body cells, like the immune system, which fight the cancer”, Tipatet said.

The molecular fingerprints allowed them to look for the very earliest signs of stage 1A breast cancer, when the growth is just a few millimeters in size and is often missed in mammograms or biopsies.

The researchers took blood samples from patients and analyzed them with a technique called Raman spectroscopy to measure molecule patterns from patient blood samples.

Raman spectroscopy is a commonly used chemical technique that is "beginning to show great potential for clinical diagnoses [of many diseases]”, said Juergen Popp, director of the Leibniz Institute of Photonic Technology, in Jena, Germany.

AI tool not ready for clinical use, yet

Based on the blood sample analysis, the researchers then trained a machine learning algorithm to detect breast cancer.

Testing blood samples from 24 patients, they found their AI algorithm could detect breast cancers with an accuracy of 90-100%, depending on the type of breast cancer.

Tipatet described the AI as "assisting the analysis” rather than replacing the humans who do the work detecting cancer.

While Popp was positive about the study's approach, he urged caution the small sample size "limits the broader generalizability of these findings”.

"Larger trials are essential for validating its clinical utility and scalability. [But] the high sensitivity and specificity achieved supports the potential for larger trials,” Popp said.

Tipatet said they already have plans "to do a larger study and see we can reproduce the findings in a larger trial.”

Detecting breast cancer early

Tipatet said the study shows Raman spectroscopy combined with AI could provide a new method for rapid and highly accurate detection of breast cancer at very early stages.

This would allow for earlier interventions and improve a patient's overall prospects, as cancers detected at stage 1A have much higher survival rates compared to those diagnosed at later stages.

Smaller tumors are easier to treat because they are localized to a specific area, making it easier to remove or target with treatments such as surgery or radiation.

Tipatet is already working with other researchers to test their approach in other types of cancers.

"We are looking into the so-called 'Big Four' cancers — lung, colorectal, prostate [and breast] cancer. These four cancers account for around 50% of the global incidence of cancer,” he said.

"We need to put in a lot of effort for early diagnosis and early screening. It would increase the quality of life and survival chances for millions of people worldwide."