A ship has run into a storm, and the captain is handing out life rings. For some people a ring will save their life; for others it will only get in the way. Chemotherapy after breast cancer surgery works much the same way: it doesn't help everyone, but doctors have no way of knowing in advance who it will help.
A new computer model called CTX takes on that problem. It draws on two sources: a microscopic photograph of the tumor, which shows what the cancer cells look like, and routine information about the patient, such as age and disease stage.
For every patient, the model works out two scenarios: what will happen if she is treated with hormones alone, and what will change if chemotherapy is added. The difference between those predictions is the benefit of chemotherapy for that particular woman. If the benefit is large, the treatment is worth prescribing; if it is small, she can do without it. This approach is more accurate than a genetic test, which merely estimates the risk of the disease returning but says nothing about whether the drug will help. As a result, chemotherapy could be prescribed 30% less often without worsening treatment outcomes.
The Falcon model used to analyze histology slides was trained on two billion tissue patches, roughly 40,000 whole slides, comparable to the archive of a major cancer center accumulated over decades.
The idea of predicting an individual's response to treatment echoes the film Gattaca, where genetic analysis determines a person's fate. Here, however, it is not about fatalism but about the ability to choose the optimal therapy based on data.