ML or machine learning is a field of scientific computing that has been around in the last few decades to explore the complex phenomena based on massive amounts of observational data. When it comes to biomedicine, an increasing amount of literature uses ML-based approaches as an alternative to regular statistical conclusions. Saar Pilosof talks about machine learning in biomedicine.
Technological development in the last years allowed collecting huge amounts of data in nearly all human fields, from economics to social interactions, passing through biomedicine. Along with the availability of such vast datasets and growing power of computers, the need emerged to take benefit of statistical frameworks capable of investigating complex data and interpret it for real-life applications.
The term machine learning contains a range of tools which try to achieve complex tasks such as detecting similarities between objects and predicting future events. The objects might be faces, individuals, weapons, cars or pathologies.
Understanding Machine Learning
The final aim of a model is to recognize the hidden connection between a set of variables having an access to (input data) and some others which are interesting for our purposes (outcomes). Samples symbolizing the displacement of the flying object are required to recognize its trajectory likewise temperature, pressure and humidity are required to predict the weather. In such conditions, we have an objective to predict future events.
And we might want to recognize, on the other hand, in a clinical trial, the sub-group of subjects that respond better to a particular treatment depending on some signs that are reported by clinicians. In such conditions, we have an objective instead to perform a classification.
In the above examples, an enormous amount of data is required to develop a robust model. Besides that, while for some situations analytical models are available, usually the nature of the relationship researched is too complex to translate them into simple formulations.
Machine learning algorithms have been designed with the aim to drive the development process of a model based on the data themselves. Technically, data are used to give training to such models to offer the most precise outcome.
A consistent group of literature has been adopting ML-based approaches in last some years for biomedical applications. ML success is actually witnessed by the number of recently published reviews from for a general medical audience to for those focused on specific subfields like neurosurgery or radiology. It’s clear that impact of ML will still increase in the coming years and will ultimately become of prime importance to public health agencies.