Overview basic types of ML algorithms

Machine learning, and types of ML algorithms, in particular, are changing the field of medicine. Thanks to their ability to analyse complex, multidimensional medical data, they enable more precise diagnosis of diseases, personalization of therapies, and optimization of clinical processes.

There are many different types of medical algorithms, and it isn’t easy to discuss them all in one post. That’s why here we would like to focus only on four of them, such as: supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning. We will present each of those approaches from both theoretical and practical perspectives in medical use.

Supervised learning algorithms use a training dataset with known outcomes to build a hypothetical function with decision variables. This function can later predict unknown samples. Researchers often use this type of algorithm to build classification models and test them to predict treatment outcomes. Examples of types of supervised algorithms:

• Naïve Bayes (NB)
• Decision tree (DT)
• Artificial neural network (ANN)
• Ensemble learning system
• Random forest (RF)
• Deep learning (DL)
• Support vector machine (SVM)
• K-nearest neighbor (KNN)

This algorithm can interpret EKGs automatically by performing pattern recognition to select from a limited set of diagnoses (i.e. a classification task). For example, in radiology, automated detection of a lung nodule from a chest X-ray would also represent supervised learning. In both these cases, the computer is approximating what a trained physician is already capable of doing with high accuracy.

This ML algorithm method learns from data without human supervision. Unlike supervised learning, unsupervised machine learning models are given unlabeled data and allowed to discover patterns and insights without any explicit guidance or instruction.

• K-means clustering
• Hierarchical clustering
• Fuzzy-c-means (FCM)

Unsupervised learning algorithms group patients based on the similarity of their medical data. This can identify new disease subtypes or discover new risk factors.

This learning method combines elements of supervised and unsupervised learning. It starts by training a supervised algorithm on a data set with known labels. An unsupervised algorithm then generates new, unlabelled data points. These new points are added to the labelled dataset, expanding the training data and potentially improving the performance of the supervised algorithm.

In medicine, semi-supervised learning algorithms can be used to analyse large medical datasets, where only part of the data is described in detail. This makes more efficient use of the available data and improves the accuracy of predictions.

A reinforcement learning is also one of the basic types of ML algorithms algorithm uses a sequential decision problem in which the algorithm solves a problem and learns from the solution. In this case, the algorithm discovers which actions produce the best results, on a trial-and-error basis. In contrast to supervised and unsupervised learning – the task of reinforcement learning is not to describe existing sets of ready-made elements, but to prepare the environment. Based on the observations, the ‘agent’ then chooses the best action (or sequence of actions) it can perform at that moment.

We can use reinforcement learning algorithms to optimize patient treatment. The algorithm can make decisions about the order in which tests are performed, the type of medication, and the dosage. It can support the development of personalized treatments in accordance with the more general precision medicine vision.

Looking at technological progress, it is certain that in the future, we can expect even more advanced applications of machine learning in medicine. Thanks to the development of deep learning algorithms, we will be able to analyse increasingly more complex, than nowadays, medical data. This, in turn, will enable us to detect diseases earlier, personalise treatments and develop new drugs. However, in order to fully exploit the potential of artificial intelligence in medicine, continuous improvement of the algorithms and close collaboration between computer scientists, mathematicians and doctors is essential.

A deeper understanding of these analytical methods can incorporate types of ML algorithms, such as unsupervised machine learning methods, into health sciences research to identify novel risk factors, improve prevention strategies, and facilitate delivery of personalized therapies and targeted patient care.

What is worth highlighting, predictive algorithms in medicine are the future of modern patients’ treatments. But they are still in the development phase. In one of our previous blog posts we raised the subject of AI for clinical prediction. In a nutshell, it explained how medicine currently uses artificial intelligence. If this topic is close to you, we highly recommend reading it.

References:

Machine Learning in Medicine:  https://pmc.ncbi.nlm.nih.gov/articles/PMC5831252/

Machine Learning (ML) in Medicine: Review, Applications, and Challenges: https://www.mdpi.com/2227-7390/9/22/2970