Machine learning supports decision making during fertility treatment
Computer scientists at the University of Southampton are using machine learning to help doctors and patients make more informed decisions during fertility treatment.
The collaboration between the University's IT Innovation Centre, and the University Hospitals Southampton Foundation Trust is personalising and streamlining the management of treatment to reduce face-to-face contact during the COVID-19 pandemic.
Analysis of past treatment records has generated reliable predictions of key fertility treatment outcomes that can be used as a decision-making support tool. The new insight can also identify which clinical appointments could be minimised while still maintaining the same level of care.
The Southampton team, led by Dr Isla Robertson and Professor Ying Cheong, is designing new prospective trials for the algorithms and seeking to change data collection practices that will help optimise future models.
Dr Francis Chmiel, an Enterprise Fellow in Electronics and Computer Science, says: "Recent developments in machine learning and data science methods mean it has become much easier to interrogate large databases of healthcare data to draw out clinical insights which could be of benefit to patients.
"In this study we have developed predictions that allow patients to be better informed about their chances of success throughout their fertility treatment cycle. By providing these predictions, under certain conditions, patients could choose a route that best suits their personal circumstances. This information can also provide more context for the clinical care team to manage patient expectations and support their wellbeing throughout their treatment cycle."
The importance of informed decision making is even greater during the current pandemic, where some patients are delaying treatment because of the potential risk of infection. This new insight could highlight cases where the chances of implantation would significantly decrease if a patient waited for a year, therefore building a more complete picture to decide on treatment dates.
The new research has also been further motivated by clinicians being asked to minimise contact with patients during the COVID-19 pandemic.
During fertility treatment patients can visit the care team up to every other day for measurements that monitor their progress and help predict when they should receive medication, Francis says. Our analysis of retrospective cycles has identified days of treatment cycles where the measurements were least predictive and therefore of least use to the clinical care team. This understanding can identify which measurements could be dropped if the clinical care team is required to have less contact with the patient.
In fact, beyond COVID-19 our results suggest that some measurements may be largely superfluous and do not add significant value to the clinical process and patient care. Clinical trials will have to be performed but if our results translate to clinical practice then measurements could be reduced, making fertility treatment more cost effective and less demanding for the patient.