Big data, machine learning and artificial intelligence: a neurologist’s guide

Abstract

Modern clinical practice requires the integration and interpretation of ever-expanding volumes of clinical data. There is, therefore, an imperative to develop efficient ways to process and understand these large amounts of data. Neurologists work to understand the function of biological neural networks, but artificial neural networks and other forms of machine learning algorithm are likely to be increasingly encountered in clinical practice. As their use increases, clinicians will need to understand the basic principles and common types of algorithm. We aim to provide a coherent introduction to this jargon-heavy subject and equip neurologists with the tools to understand, critically appraise and apply insights from this burgeoning field.

Artificial Intelligence in Colorectal Cancer Screening, Diagnosis and Treatment. A New Era

“…Furthermore, without sufficient high-quality training data and robust computational infrastructure, even the most state-of-the-art algorithms are doomed to failure [172]. In general, deep learning algorithms require large volumes of data to train the AI systems at the best level to obtain the best results.…”

“…Despite the efforts to standardize patient care, the variability between patients concerning differences in clinical presentation and desired outcomes should be thoroughly considered prior to AI model development [173]. Any potential deviation from the training conditions may result in the unpredictable behavior of an algorithm [172].…”

“…While we know the input and output data an algorithm produces, there is limited information about its internal workings and processes. This issue is referred to as the "black box" problem in machine learning [172]. The "black box" problem indicates that significant confounding from any of the many factors could easily be missed, especially in deep learning algorithms with many hidden layers (e.g., classifying skin lesions as malignant or benign [174]).…”

Machine Learning Use for Prognostic Purposes in Multiple Sclerosis

“…In both cases ML models do not generalize to new data, thus leading to large errors on the test set, limiting the usefulness of the study. Further explanations of these terms have been recently provided, with examples, in the field of neurology [25], while some widely-used methods are explained in Section 5.…”

“…Inspection of the results of a PubMed search with keywords "multiple sclerosis" AND ("machine learning" OR "artificial intelligence" OR "neural network") retrieved 286 studies, of which eight used clinical data to derive predictions on the course of MS in individual patients. The papers are listed in Table 1, together with one work identified among the references cited by another review [25]. No additional studies were found among the first 100 results (ranked by relevance) of an identical search performed on Google Scholar, which yielded 3980 results.…”