AI News, Artificial intelligence, bias and clinical safety

Artificial Intelligence and Machine Learning in Otolaryngology

AI would mine prior data and identify which test provides the most clinically and cost-effective benefit, and would use real patient data, in real time, to constantly upgrade its evidence-based recommendations.

“By analyzing how similar patients have responded to past treatments, ML can provide information based on many more patient experiences than any individual physician could incorporate into their medical decision making,” Dr. Bur said.

According to recent estimates, for every hour that physicians provide face-to-face clinical care to patients in the outpatient setting, they spend nearly two additional hours on EHR documentation and desk work (Ann Intern Med.

“By recording and automatically extracting content from clinical encounters using natural language processing, virtual scribes have the potential to reduce the burden of clinical documentation,” Dr. Bur said.

Moberly, MD, an assistant professor of otolaryngology-head and neck surgery at The Ohio State University Wexner Medical Center in Columbus, and colleagues have used AI to develop software called “Auto-Scope” to help clinicians perform more accurate diagnoses for the ear using digital otoscopy.

“This initially involved classifying images as normal or abnormal, but will eventually expand to build enhanced composite images and provide more specific information on particular types of pathology,” he said.

Another area Dr. Friedland has studied is using natural language processing, a form of AI that interprets natural speech or written documentation, to identify expert descriptors of specific vestibular conditions.

More recently, they used a ML model known as a random forest to demonstrate that the mucus cytokines IL-5 and IL-13 were predictive of baseline olfactory function in chronic sinusitis patients.

By analyzing how similar patients have responded to past treatments, machine learning can provide information based on many more patient experiences than any individual physician could incorporate into their medical decision making —Andrés Bur, MD Elizabeth A.

Blair, MD, professor of surgery in the section of otolaryngology-head and neck surgery at the University of Chicago, pointed out that using AI and ML requires the input and acumen of expert clinicians to determine where gaps exist and what information would help physicians.

This could assist physicians in selecting the type of treatment for patients, finding alternative therapies that could be effective, or predicting and determining if they are at increased risk for complications for a certain treatment.

We need research funding to perform this work and to incentivize young physicians to study AI and ML in otolaryngology practice.” Dr. Chowdhury added that there is a large learning curve needed to truly understand the scope of AI and ML research, as it combines advanced elements of statistics, mathematics, computer science, and probability theory, with which most physicians are unfamiliar.

As CMS and private insurers continue to explore alternative payment models, it is likely that incentives for health systems and EHR developers will shift in favor of technologies—such as AI—with the potential to improve healthcare quality and cost.

In launching the AI-driven clinical data registry Reg-ENT, which is designed to harness the power of data to guide the best otolaryngology care, the American Academy of Otolaryngology–Head and Neck Surgery is taking an important step to prepare for alternative payment models and increased requirements for quality reporting that will affect reimbursement.

By participating in clinical data registries such as Reg-ENT and by seeking to collaborate with data scientists in developing new AI, otolaryngologists can ensure the field is a pioneer in the development and deployment of AI technologies in clinical practice.

For clinicians interested in being actively involved in AI and ML research, he recommended reviewing the principles behind the statistical models like linear regression and logistic regression, as these are often simpler (and better) options for predictive analytics than non-linear models.

CA: A Cancer Journal for Clinicians

The research community has yet to reach a consensus on specific data sets that can be used for comparing and contrasting efforts in terms of performance, generalizability, and reproducibility, although the volume of medical data being made public is an encouraging move forward.206 Furthermore, access to available data sets should be improved to promote intellectual collaboration.

Institutional, professional, and government groups should be encouraged to share validated data to support the development of AI algorithms, which requires overcoming certain fundamental technical, legal, and perhaps ethical concerns.207 For example, the National Institutes of Health recently shared chest x‐ray and CT repositories to help AI scientists.208 Such efforts bear expansion to a much wider audience across disease states.

Although the current state of research has prioritized performance gains over explainability and transparency, the interpretability of AI is an active area of research.209 The benefits of trust and transparency in AI systems will differ based on their performance, allowing for the identification of failures when AI is subhuman and, consequently, transforming superhuman AI into a learning resource.

Finally, automated systems also might challenge the dynamics of responsibility within the doctor‐patient relationship, as well as the expectation of confidentiality.212 In terms of regulatory aspects, the US Food and Drug Administration has been regulating automated clinical decision‐making systems since the 1990s.213 With the advent of new prediction techniques, including deep learning, predictive models seeking approval must be further scrutinized in terms of the ground truth data used in training them, their intended use cases, and their generalizability and robustness against edge cases, as well as their life‐long learning aspects, as they are continuously updated with more learning and more data.

Current cyber security research starts to offer solutions, including cryptonets, in which homomorphic encryption allows neural networks to run training and inference on encrypted data.215 Today’s diagnostic paradigm in medicine focuses on the detection of visually recognizable findings that suggest discrete pathologies in images.

It is likely that, during the early phase of AI, when human experts will continue to play key roles in gatekeeping AI’s output, the majority of incidental findings detected by AI will still be evaluated by humans to discern whether or not they are clinically significant in the same manner as when humans detected incidental findings.

This shows initial promise in several disease conditions but requires additional proof of clinical utility in prospective trials and education of physicians, technologists, and physicists to incorporate into widespread use.216, 217 Although there likely will always be a “black box” for human experts in viewing AI‐generated results, data visualization tools are increasingly available to allow some degree of visual understanding of how algorithms make decisions.127 The curation of comprehensive data sets and outcomes that incorporate both disease‐related and unrelated elements also will help train and expand AI systems to account for risks beyond cancer itself.

Conversely, strategies that predict outcomes without a ground truth provided by human experts may disrupt the traditional workflow familiar to clinicians and patients today.218 Furthermore, the increased incorporation of AI in monitoring health resources and outcomes likely will improve efficiency and reduce cost.

Data Science and AI in Pharma and Healthcare (CXOTalk #275)

Data, artificial intelligence and machine learning are having a profound influence on healthcare, drug discovery, and personalized medicine. On this episode ...

The Ethics and Governance of AI opening event, February 3, 2018

Chapter 1: 0:04 - Joi Ito Chapter 2: 1:03:27 - Jonathan Zittrain Chapter 3: 2:32:59 - Panel 1: Joi Ito moderates a panel with Pratik Shah, Karthik Dinakar, and ...

Ethics of AI

Detecting people, optimising logistics, providing translations, composing art: artificial intelligence (AI) systems are not only changing what and how we are doing ...

Algorithmic Accountability: Designing for Safety | Ben Shneiderman || Radcliffe Institute

Vital services such as communications, financial trading, health care, and transportation depend on sophisticated algorithms. Some rely on unpredictable ...

The Cognitive Era: Artificial Intelligence and Convolutional Neural Networks

Jim Hogan, Raik Brinkmann, James Gambale Jr., Chris Rowen and Drew Wingard discuss artificial intelligence and convolutional neural networks at the Diaz ...

AI Now 2017 - Experts Workshop

The Robot Will See You Now: U of T experts on the revolution of artificial intelligence in medicine

Make room, stethoscope and otoscope. Artificial intelligence (AI) applications are increasingly among the physician's standard instruments, experts at the ...

AI for Good Panel - Precision Medicine | SXSW 2017

Fiona Godlee: Panel discussion on "Low carb high fat diets: Public controversies and opportunities"

Fiona Godlee, Editor in Chief, The BMJ, moderated the panel discussion on "Low carb high fat diets: Public controversies and opportunities" at the Swiss Re ...

Information - The latest news in Healthcare IT – straight to your inbox.

Information - The latest news in Healthcare IT – straight to your inbox. Despite being touted as next-generation cure-alls that will transform healthcare in ...