Putting it all together – specifying a use case

Now that we've seen some of the ways in which machine learning problems can vary in healthcare, it becomes easier to specify a problem. Once you've selected a population, a medical task, an outcome measure, and disease, you can formulate a machine learning problem with a reasonable amount of specificity. We haven't included our choice of an algorithm in our discussion because, technically, it is separate from the problem being solved, and also because many problems are approached by using multiple algorithms. We will look at specific machine learning algorithms in Chapters 3 and 7, which will provide you with some background for choosing algorithms.

Here are some example use cases that can be specified using the preceding information:

"I'd like to predict which healthy elderly adults are likely to be diagnosed with Alzheimer's disease in the next five years."

"We are going to build a model that looks at images of moles and predicts whether the moles are likely to be benign or malignant."

"Can we predict whether pediatric patients presenting to the emergency room with asthma will be admitted to the hospital or discharged home?"