September 3, 2019
2 minute read
Click here to read the related cover story, “Artificial intelligence is impacting kidney disease treatment.”
AI is an umbrella term that brings together concepts from several fields, including statistics, algorithms, machine learning, information retrieval, and data science in general. Machine learning (ML), a subfield of AI, is the scientific study of algorithms and statistical models that can learn how to perform specific tasks by relying on patterns and inference without the use of explicit instructions. As a result, ML algorithms are very “data-hungry” and often require thousands of observations to reach acceptable performance.2
The vast amount of data collected in electronic medical records at the point of care provides a rich platform for adopting ML. It is excellent at handling large numbers of predictors and can combine predictors in a nonlinear and highly interactive way. This capability enables the exploitation of new types of data (such as free text, images, audio, and temporal data) that were previously impossible to analyze due to their sheer volume or complexity.
Deep learning (DL) is another subset of ML that uses multilayer artificial neural networks (ANNs) to create more sophisticated nonlinear feature engineering than traditional ML techniques (see Figure 2).3
A recent bibliographical study on the global evolution of AI research in healthcare and medicine found that AI technology and clinical applications of AI are relatively common in professions such as ophthalmology, oncology, and cardiology. it is clear.Four Although the daily disease burden for patients with CKD and ESKD is high compared to other diseases, it is less prevalent in nephrology.Five
Source: Fresenius Medical Care
In ophthalmology, multiple AI-based grading algorithms have been developed to screen fundus photographs obtained from diabetic patients and identify who should be referred to an ophthalmologist for treatment. The developed DL model has high reliability with a sensitivity of 94%–98% and specificity of 93%–98% across different models.6,7
The first AI-based device approved by the FDA and cleared for sale was IDx-DR (IDx Technologies Inc., Coralville, Iowa). The device combines DL image recognition software to analyze retinal images and provide recommendations for referring patients for evaluation. Diabetic retinopathy.8 In oncology, AI has aided tumor imaging, pathology, and clinical decision-making.9
AI also enters patients’ lives in the form of wearable devices, remotely monitoring patients and providing personalized recommendations. Current Health's AI-enabled wearable device measures multiple vital signs and recently received FDA clearance for patients to use at home.Ten The device can measure pulse, respiration, oxygen saturation, body temperature, and motility. The device provides patients and doctors with real-time updates and alerts on vital signs, making patient treatment easier and allowing doctors to quickly address complications. This technology uses ML to analyze continuous vital signs and dynamic data to detect abnormal trends.
Currently, this device is primarily used for patients with chronic obstructive pulmonary disease and heart failure. AI-powered remote monitoring could help improve hospitalization and mortality rates for such patients.11
