Artificial Intelligence in Clinical Health Care Applications: Viewpoint

Abstract

The idea of artificial intelligence (AI) has a long history. It turned out, however, that reaching intelligence at human levels is more complicated than originally anticipated. Currently, we are experiencing a renewed interest in AI, fueled by an enormous increase in computing power and an even larger increase in data, in combination with improved AI technologies like deep learning. Healthcare is considered the next domain to be revolutionized by artificial intelligence. While AI approaches are excellently suited to develop certain algorithms, for biomedical applications there are specific challenges. We propose six recommendations—the 6Rs—to improve AI projects in the biomedical space, especially clinical health care, and to facilitate communication between AI scientists and medical doctors: (1) Relevant and well-defined clinical question first; (2) Right data (ie, representative and of good quality); (3) Ratio between number of patients and their variables should fit the AI method; (4) Relationship between data and ground truth should be as direct and causal as possible; (5) Regulatory ready; enabling validation; and (6) Right AI method.

Estrogen Receptor Pathway Activity Score to Predict Clinical Response or Resistance to Neoadjuvant Endocrine Therapy in Primary Breast Cancer

“…In brief, data-driven efforts to generate pathway gene expression signatures, such as gene set enrichment analysis and DAVID, have been typically based on discovery of noncausal associations between expressed genes and a pathway, by analyzing datasets with more or less specified relationships to the pathway (49,50). Such methods are prone to finding spurious associations and carry a high risk of overfitting, which interferes with biological validation and performance on independent datasets and with adaptation to other gene measurement platforms (51). In contrast, the Bayesian ER pathway model was biologically validated, can calculate the ERPAS of an individual sample despite variations in the expressed target gene subset, and its performance is relatively independent of mRNA measurement platform used.…”

Health Care Employees’ Perceptions of the Use of Artificial Intelligence Applications: Survey Study

“…Recently, health care systems in several countries have begun to rely on storage of patient information to provide the best quality of health care. Due to rapid technological developments, health care information technology solutions provide the capacity to store enormous volumes of patient data; however, appropriate utilization of this data is essential to enhance health care quality, improve decision making, and reduce costs [ 1 , 2 ]. Over the last decade, artificial intelligence (AI) has provided significant advancements in this regard [ 3 ].…”

Age Assessment of Youth and Young Adults Using Magnetic Resonance Imaging of the Knee: A Deep Learning Approach

“…The use of machine learning for health applications is not new and is broadly employed for disease prediction and prognosis [14,15], genomics, proteomics, and microarrays [16]; it has also been used to predict health care utilization through Web search logs [17]. Contrary to many machine learning techniques, deep learning methods perform feature engineering: instead of having a domain expert specify important data characteristics, it learns the informative representations in the data and performs a task of classification or regression [18,19]. When working with medical images, this is especially advantageous since image features are difficult to translate into descriptive means [20].…”