Researchers in the field of artificial intelligence aim to find ways for computers to respond to questions and make personalised recommendations. This goal becomes more tractable when applied to a specific area of expertise. I have been involved with a team that has developed an artificial intelligence system to help women understand the variation in their female hormones.
The hormone networks that control female menstrual cycles represent the most complex aspect of the endocrine system. Intricate feedback pathways between the hypothalamus, pituitary and the ovaries trigger a succession of fluctuations in hormone levels resulting in ovulation. Although healthy women should expect these cycles to repeat regularly for 30 to 40 years, female hormones are sensitive to external stress factors, such physical exercise, suboptimal nutrition, disrupted sleep or mental pressures. These cause variations between cycles. Furthermore, female hormones are picked up by receptors all around the body, including brain, bone and heart cells. For some women, these effects are more troublesome than for others.
Barometer of health
In addition to being essential for fertility, a well-functioning menstrual cycle provides women with a more general barometer of health. Some women try to track their hormonal fluctuations using proxy measurements, like body temperature or secondary metabolites in saliva or urine. But the gold standard is to determine the actual serum levels of hormones circulating in the blood. A pin-prick sample can measure all four key female hormones: oestradiol, progesterone, follicle stimulating hormone (FSH) and luteinising hormone (LH).
Although taking a blood sample every day would be troublesome and expensive, measuring the hormones in just two pin-pricks over a month is certainly manageable. This is where Bayesian inference can help.
In his brilliant book, Probability Theory: The Logic of Science, Edwin Jaynes explains the power of the Bayesian approach. It provides the fundamental way to draw statistical inferences from data, based on our prior knowledge. In the case of hormone intelligence, science provides a good understanding of the endocrine networks and the degree of variation seen across the population. If we are limited to just two measurements per month, days 14 and 21 provide the most information because these are likely to be close to the main hormone peaks that occur over a cycle.
For any particular woman, Bayesian inference can be used to select the most likely sets of hormone curves that are consistent with her blood test results and other information she has provided.
A clever mathematical algorithm that draws a set of hormone curves is not much use to a woman without the interpretation and advice of a specialist endocrinologist. But special expertise is limited and expensive. This is where an expert system can help.
Computer systems that emulate human expertise have been around for nearly 50 years. I had the good fortune to work closely with Dr Nicky Keay in order to ensure that the system acquired sufficient knowledge to generate personalised, expert reports. In a process reminiscent of Borges, we explored the tree of all possible paths, giving an interpretation for each outcome. Dr Keay was insistent that all advice was backed up by appropriate evidence and scientific references.
One of the most interesting aspects of female hormone profiles is that they change as a woman approaches menopause: the ovarian hormones decline, whereas the control hormones become elevated. This allowed us to create an innovative score to evaluate a woman’s ovarian responsiveness. The score was designed to be particularly useful for women in the perimenopausal stage of life.
Mapping Female Hormones
Combining Bayesian inference with an expert system has created a genuine artificial intelligence system. It takes as inputs, blood test results and self-reported data, including personal wellbeing scores, and it generates an expert report with personalised advice.
I hope that mapping female hormones will empower women with a better understanding of their personal hormones and help improve their wellbeing.