Machine learning (overview)

Andrew Murphy et al.

Machine learning is an avenue of computer science that can extrapolate information based on observed patterns. More specifically, machine learning is becoming an increasingly important tool in the medical profession as a primary computer-aided diagnosis algorithm or a decision support system. It is highly likely that in the next 10-20 years various implementations of machine learning will have a very profound impact on the way radiology is practised and it seems, at least to many in the field, inevitable that many of the tasks that are currently considered core to the practice of radiology (e.g. abnormality detection and classification) will be performed at least in part by these systems. It is therefore prudent that radiologists become familiar with the fundamentals of these approaches. 

Although there are countless specific models and implementations of machine learning, the majority fall into one of a relatively small number of fundamentally different underlying learning processes and models. 

Learning processes

The specifics of how a machine learning algorithm is trained to recognise certain features and thereby become able to make accurate predictions on new examples varies depending on the type of data being used and the algorithm architecture. Four of the most commonly used learning processes are:

  1. supervised learning
    • most commonly applied in radiology
    • a set of images (training set) is provided along with a known feature or diagnosis (ground truth)
    • machine learning algorithm makes predictions based on response values
    • because the input data (training set) and the response values are already known (ground truth) the algorithm can make iterations until it reaches an agreed-upon result
  2. unsupervised learning
    • less common in radiology
    • algorithm is only fed input data with no known ground truth
    • often used to identify trends or patterns within a data
  3. reinforcement learning
    • applied in the context of an agent inspecting its surroundings and performing actions in order to receive delayed rewards
    • even less commonly used in radiology but recently used in image analysis tasks for high-resolution images in order to focus on inspecting the relevant areas 
  4. evolutionary algorithms
    • a population-based metaheuristic optimization algorithm used in evolutional science


How the aforementioned learning processes are implemented is variable and determined in part by the type of problem being solved. Although much of the recent work in the field of image processing generally, and more specifically radiology, has focused on convolutional neural networks, a type of neural network, a number of other models are useful in various circumstances. These include: 

  • linear regression
  • logistic regression
  • decision tree
  • random forest
  • support vector machine
  • neural networks
Machine learning
Share article

Article information

rID: 56095
Synonyms or Alternate Spellings:

Support Radiopaedia and see fewer ads

Updating… Please wait.

Alert accept

Error Unable to process the form. Check for errors and try again.

Alert accept Thank you for updating your details.