Article Text

Download PDFPDF
What is machine learning?
  1. Orkun Baloglu1,2,
  2. Samir Q Latifi1,2,
  3. Aziz Nazha3
  1. 1 Department of Pediatric Critical Care Medicine, Cleveland Clinic Children’s, Cleveland Clinic, Cleveland, Ohio, USA
  2. 2 Cleveland Clinic Children's Center for Artificial Intelligence, Cleveland, Ohio, USA
  3. 3 Department of Medical Hematology and Oncology, Taussig Cancer Institute, Cleveland Clinic, Cleveland, Ohio, USA
  1. Correspondence to Dr Orkun Baloglu, Department of Pediatric Critical Care Medicine, Cleveland Clinic Children’s, Cleveland Clinic, Cleveland, Ohio, USA; BALOGLO{at}ccf.org

Statistics from Altmetric.com

The term machine learning (ML) is emerging more often in the medical literature. There are successful clinical applications of ML with the specialties of ophthalmology and radiology leading the way. For example, in ophthalmology the diagnosis of diabetic retinopathy and retinopathy of prematurity,1 and in radiology the diagnosis of stroke or cancers from digital images is promising.2 This success is expected to expand into other medical disciplines including general paediatrics and its subspecialties. Therefore, healthcare practitioners and researchers are most likely to benefit from getting familiar with ML terminology in order to stay up to date and better understand the ML research methodology and its applications in medicine. This brief review is intended to be an introduction to ML and its associated terminology, and to review selected studies with ML models from the paediatric literature.

What is machine learning?

Predictive models are mathematical functions that take input variable(s), process the input variable(s) and produce an output. The mathematical definition of processing the input variable(s) to produce an output is an algorithm. Describing the relation between the variables in a dataset can be feasible by linear equations in some datasets. For instance, linear regression models aim to fit a linear line that relates input variable(s) to output variable(s). However, when there are a large number of variables, or their relation to the output variable is too complex, or non-linear to be described by linear lines, or even there is no obvious output variable to be predicted in the dataset, then the classic linear regression models fail. In such cases, a different approach is needed. This alternative strategy is called ML. In ML, algorithms are developed to discover unknown complex relations between …

View Full Text

Footnotes

  • Twitter @BalogluMD, @AzizNazhaMD

  • Contributors OB wrote the draft of the manuscript. OB, SQL and AN revised and finalised the manuscript prior to submission. OB prepared the figure. All authors approved the final manuscript and figure.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Provenance and peer review Commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.