Post 1

Machine Learning: what is it and how is it applied to the healthcare field? Challenges.

The machine learning is a subset of artificial intelligence (AI) that involves using algorithms to learn from data and make predictions based on that data.1

In the case of healthcare, clinical machine learning (ML) data comes in various forms: radiological scans, electrocardiograms, clinical events organized in time series and clinical notes among others. 2

It is common for ML healthcare models not to be generalizable, as many are developed in highly controlled environments, using sensitive data curation methods, but are generally expected to perform well when exposed to the volatility of real-world data. (2)

Examples of real-world data that contribute to the generation of volatility include: measurement drift due to changes in hardware (e.g., use of a Philips versus a General Electric scanner), different clinical conditions (e.g., a metropolitan area hospital versus a rural clinic), changes in population health (e.g., increasing rates of diabetes), global health challenges (e.g., a pandemic), economic disparities (e.g., access to health care), and many others.(2)

If these real-world data scenarios are not taken into account, model performance will progressively decrease or fail due to data streams never conceived during training. (2)

For this reason, it is critical and the foundation of any real-world machine learning system to perform proper data collection, cleaning and validation; without it, patient care will suffer from automation and poor performance.

At Datatherapy, we help stakeholders in the healthcare ecosystem make health impact decisions based on the power of data.