Vaginal swab test using AI found to accurately detect endometrial cancer
Noninvasive test called a 'breakthrough' in gynecologic cancer diagnosis
Written by |
A test that uses artificial intelligence (AI) to analyze genetic data from a vaginal swab can detect endometrial cancer with high diagnostic accuracy, a new study reports.
According to Pinkdx, the company that developed the test, the use of AI combined with a noninvasive swab marks a “breakthrough [that] opens the door to a radically less invasive diagnostic pathway for women.” As of now, this type of gynecological cancer typically requires more invasive testing to detect.
“This peer-reviewed publication provides clear evidence that endometrial cancer signals can be detected from a vaginal swab. It represents a scientific breakthrough and a critical step toward a diagnostic pathway designed around women — not procedures,” Bonnie Anderson, cofounder, chair, and CEO of Pinkdx, which funded the study, said in a press release announcing its publication.
A peer review means a journal article is vetted by experts in the same field to ensure the accuracy and validity of the findings being reported.
The study, “Whole-transcriptome sequencing and machine learning detect molecular signatures of endometrial cancer in non-invasive vaginal swabs,” was published in the International Journal of Gynecological Cancer.Â
Endometrial cancer, which is marked by the abnormal growth of cells that line the inside of the uterus, is the most common form of gynecological cancer. It is considered highly treatable if it’s found early.
Investigating vaginal swabs as a noninvasive diagnostic tool
Abnormal vaginal bleeding is a common symptom of this cancer type, so women who experience unexplained bleeding are encouraged to get tested. But currently available tests to diagnose endometrial cancer are quite invasive, usually requiring the collection of tissue from the uterus.
The gynecologic cancer diagnostic journey is too invasive, too uncertain, and too burdensome for women.
Moreover, because abnormal vaginal bleeding can also be caused by a host of other conditions, such invasive procedures may be a little bit like using a water cannon to hunt a mosquito. Statistics show that nearly 90% of people who undergo such testing end up not having cancer.
“The gynecologic cancer diagnostic journey is too invasive, too uncertain, and too burdensome for women,” Anderson said.
This new study aimed to challenge a longstanding belief: that to make meaningful conclusions about cell activity in the uterus, it’s necessary to obtain a sample of cells from the uterus itself.
The researchers theorized that it may be possible to detect signs of endometrial cancer by analyzing cells collected from the vagina via a much less invasive swab.
Giulia Kennedy, PhD, cofounder and chief scientific officer of Pinkdx, noted that “the assumption has always been that you have to sample the uterus to understand what’s happening in the uterus.”
“Our findings challenge that assumption. We show that molecular signals associated with endometrial cancer can be detected from a vaginal swab — a result with meaningful implications for how we rethink the diagnostic experience for women,” Kennedy said.
Researchers used AI alogrithm to detect patterns indicating cancer
To test this idea, the researchers used data from about two dozen women who underwent hysterectomy as part of a now-completed clinical trial (NCT6460818). Because these women had their uteruses removed, the researchers were able to analyze data from vaginal swabs and compare the findings against the gold standard of uterus tissue. Data from publicly-available genetic datasets were also used to supplement the analyses.
Samples were analyzed using a technique called whole-transcriptome RNA sequencing. When a gene is read, the genetic code is copied over from a cell’s DNA into another molecule called RNA, which may then be used as a template to make protein or serve other functions in the cell. Whole-transcriptome RNA sequencing essentially determines the sequences of all RNA molecules in a sample, providing a snapshot of the cells’ genetic activity.
These sequencing data were then analyzed using AI algorithms to detect patterns indicative of endometrial cancer, according to the researchers.
To evaluate the accuracy of their test, the team used a statistical metric called the area under the curve (AUC), which measures how well a test distinguishes between two groups — here, patients with or without endometrial cancer. AUC values range from 0.5, meaning the test is no better than random chance, to one, which indicates that the test is perfect at differentiating the groups.
The data here indicated that whole-transcriptome RNA sequencing data from noninvasive vaginal swabs could detect endometrial cancer with an AUC of up to 0.98, suggesting nearly perfect accuracy.
These findings suggest that “diagnostic testing from a minimally invasive vaginal swab, combined with clinical judgment and follow-up, could potentially reduce the need for invasive tissue sampling among women predicted to be at low risk of harboring endometrial cancer,” the scientists wrote.
The researchers stressed that these analyses were limited to a small number of patients and noted that additional tests are needed to further validate this approach. Pinkdx said it is already advancing multiple validation studies to test this approach in broader, real-world populations.
“Our focus is not just discovery, but translation — designing molecular diagnostics that can move into real-world clinical decision-making. That’s the lens through which this work was conceived,” Kennedy said.
