Woman joins AI research program by chance, is shocked when it discovers something the human eye cannot detect
The good side of AI.
An experimental artificial intelligence system recently helped identify breast cancer early in Yvonne Cook, a woman who only joined the AI research program by chance. Yvonne’s case highlights how AI can spot things the human eye just can’t detect, offering a hopeful glimpse into the future of medical technology.
Yvonne shared her remarkable story, explaining how she was called for a routine appointment on Thursday, March 26. According to LADBible, she noticed information about an AI research program currently running. A big poster on display reminded her that she could opt out, but she told herself, “I’ve got nothing to lose.” She figured it’s always good to take part in medical research when you can, so she stuck on.
Soon after, Yvonne received a recall letter. She described it as pretty vague and gentle, suggesting that perhaps her initial mammogram had been difficult to read. Her follow-up appointment was with Dr. Gerald, who explained that the AI system used in screening her mammogram had flagged something undetectable by human eyes.
Yvonne was more than grateful to receive an early diagnosis
A scan confirmed her diagnosis: a small, grade two tumor. It was caught at a very early stage, and a biopsy later confirmed the cancer type. Yvonne felt incredibly grateful that her mammogram coincided with this research period and that AI had picked up on it, maintaining a positive mindset throughout her treatment.
Dr. Gerald, who also spoke about the case, emphasized that using AI can improve breast cancer detection by more than 10%. He made it clear that human doctors and scientists are still very much involved in the process, working alongside the AI. It’s not about replacing people, but augmenting their abilities. This isn’t the only way AI is shaking things up for cancer patients.
New research from the University of Geneva suggests we might soon be able to track how cancer spreads. This could lead to a huge shift in treatments. Imagine the possibility of personalized treatments that could stop illnesses from metastasizing, or forming new, secondary tumors in other parts of the body.
Ariel Ruiz i Altaba, a professor at the UNIGE Faculty of Medicine and an author of the study, is interested in why some cancer cells break away and migrate while others stay put. He pointed out that the big challenge is understanding the logic of cancer and identifying the characteristics of cells that separate from a primary tumor to create another one elsewhere. The difficulty is that analyzing a cell’s complete molecular identity destroys it, but observing its function requires it to be alive.
To get around this, the researchers isolated, cloned, and grew tumor cells in the lab. They then analyzed the gene signatures from thirty different cell clones. This work helped them discover that metastatic potential isn’t determined by a single cell’s profile.
Instead, it’s all about how groups of related cancer cells interact with each other. They fed these findings into a specialized AI program called Mangrove Gene Signatures (MangroveGS). After training, this AI was able to predict metastasis and colon cancer recurrence with nearly 80% accuracy.
The hope is that MangroveGS could soon work with doctors in hospitals. It could provide a metastasis risk score to health professionals and patients, helping determine the intensity of treatment. Since chemotherapy and radiotherapy essentially involve poisoning the body to kill cancer cells, this information is crucial.
Professor Altaba believes this will prevent overtreatment for low-risk patients, cutting down on side effects and unnecessary costs. At the same time, it would intensify monitoring and treatment for those at high risk. It also offers a fantastic way to optimize participant selection for clinical trials, reducing the number of volunteers needed and boosting the statistical power of studies, ultimately benefiting the patients who need it most.
Have a tip we should know? [email protected]
