The discovery of new causal links in existing medical data could eliminate many a clinical trial.

The correct diagnosis of a disease depends on the knowledge of which medical conditions cause what symptoms. Similarly, its correct treatment depends on knowing a drug’s effects or those from a change in lifestyle. Therefore, the knowledge of “cause and effect” is essential in the medical world.

However, there is a vast store of data from medical observational studies or randomized controlled trials accumulated over the decades. What if we could deploy an AI algo to scour this data for new illumination on “cause and effect”? Get fresh insight on symptoms and the likely disease?

But a unidimensional examination of correlation inside a single data set (say, obesity and heart disease), or that in another data set (low vitamin D and obesity), will not tell us the answer to the causal relationship between low vitamin D and heart disease.

What’s needed is a multi-dimensional ability – an AI tool to scour for causal nexus across many and different data sets.

Cause and effect: Work by researchers Anish Dhir and Ciarán Lee may be the answer

Dhir and Lee, who work at Babylon Health, a digital health care provider in the UK, have incorporated a technique from quantum cryptography to address the challenge. (Lee also works on machine learning and quantum computing at University College, London.)

Their algorithm can extract the maximum information on cause and effect from multiple data sets that share certain common variables and have been combined for the analysis.

For example, they tested the system on datasets where the causal nexus was already identified, such as measurements of the size and texture of breast tumors. The algo correctly pinpointed the links between the size, texture, and malignancy of the tumors.

Their work has been peer-reviewed. It will be presented at the Association for Advancement of Artificial Intelligence conference in New York this week.

Related Story:     A.I. Algorithms Predict the Likelihood of Breast Cancer