Toward Generalizability in the Deployment of Artificial Intelligence                     in Radiology: Role of Computation Stress Testing to Overcome                     Underspecification

Eche, Thomas; Schwartz, Lawrence H.; Mokrane, Fatima-Zohra; Dercle, Laurent

doi:10.1148/ryai.2021210097

Cited by 82 publications

(37 citation statements)

References 72 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“… 16 Lastly, model development will need to implement stress testing to address the issue of underspecification, or failing to capture the inner logic of an underlying system due to confounding factors in data distributions. 27 To ensure broad generalizability, these stress tests should become standard practice, just as crash tests are fundamental to the automotive industry.…”

Section: Discussionmentioning

confidence: 99%

Artificial intelligence and radiomics: fundamentals, applications, and challenges in immunotherapy

Dercle

McGale

Sun

et al. 2022

J Immunother Cancer

Self Cite

View full text Add to dashboard Cite

Immunotherapy offers the potential for durable clinical benefit but calls into question the association between tumor size and outcome that currently forms the basis for imaging-guided treatment. Artificial intelligence (AI) and radiomics allow for discovery of novel patterns in medical images that can increase radiology’s role in management of patients with cancer, although methodological issues in the literature limit its clinical application. Using keywords related to immunotherapy and radiomics, we performed a literature review of MEDLINE, CENTRAL, and Embase from database inception through February 2022. We removed all duplicates, non-English language reports, abstracts, reviews, editorials, perspectives, case reports, book chapters, and non-relevant studies. From the remaining articles, the following information was extracted: publication information, sample size, primary tumor site, imaging modality, primary and secondary study objectives, data collection strategy (retrospective vs prospective, single center vs multicenter), radiomic signature validation strategy, signature performance, and metrics for calculation of a Radiomics Quality Score (RQS). We identified 351 studies, of which 87 were unique reports relevant to our research question. The median (IQR) of cohort sizes was 101 (57–180). Primary stated goals for radiomics model development were prognostication (n=29, 33.3%), treatment response prediction (n=24, 27.6%), and characterization of tumor phenotype (n=14, 16.1%) or immune environment (n=13, 14.9%). Most studies were retrospective (n=75, 86.2%) and recruited patients from a single center (n=57, 65.5%). For studies with available information on model testing, most (n=54, 65.9%) used a validation set or better. Performance metrics were generally highest for radiomics signatures predicting treatment response or tumor phenotype, as opposed to immune environment and overall prognosis. Out of a possible maximum of 36 points, the median (IQR) of RQS was 12 (10–16). While a rapidly increasing number of promising results offer proof of concept that AI and radiomics could drive precision medicine approaches for a wide range of indications, standardizing the data collection as well as optimizing the methodological quality and rigor are necessary before these results can be translated into clinical practice.

show abstract

Section: Discussionmentioning

confidence: 99%

Artificial intelligence and radiomics: fundamentals, applications, and challenges in immunotherapy

Dercle

McGale

Sun

et al. 2022

J Immunother Cancer

Self Cite

View full text Add to dashboard Cite

show abstract

“…Medical imaging research is facing a proliferation of studies proposing new AI-powered diagnostic and predictive tools, and promising outstanding performances. Relatively few studies have focused on evaluating the applicability of these models, and what concrete benefits they would bring to clinical practice in the real-world setting [ 110 , 111 , 112 ]: the results can be surprisingly disappointing, when such models are tested in different cohorts of patients or institutions. In the emergency setting, for example, disappointing results achieved by some software should warn radiologists against placing excessive reliance on these tools 90 .…”

Section: Challenges and Perspectivesmentioning

confidence: 99%

Artificial Intelligence in Emergency Radiology: Where Are We Going?

et al. 2022

View full text Add to dashboard Cite

Emergency Radiology is a unique branch of imaging, as rapidity in the diagnosis and management of different pathologies is essential to saving patients’ lives. Artificial Intelligence (AI) has many potential applications in emergency radiology: firstly, image acquisition can be facilitated by reducing acquisition times through automatic positioning and minimizing artifacts with AI-based reconstruction systems to optimize image quality, even in critical patients; secondly, it enables an efficient workflow (AI algorithms integrated with RIS–PACS workflow), by analyzing the characteristics and images of patients, detecting high-priority examinations and patients with emergent critical findings. Different machine and deep learning algorithms have been trained for the automated detection of different types of emergency disorders (e.g., intracranial hemorrhage, bone fractures, pneumonia), to help radiologists to detect relevant findings. AI-based smart reporting, summarizing patients’ clinical data, and analyzing the grading of the imaging abnormalities, can provide an objective indicator of the disease’s severity, resulting in quick and optimized treatment planning. In this review, we provide an overview of the different AI tools available in emergency radiology, to keep radiologists up to date on the current technological evolution in this field.

show abstract

“…The reasons behind the translational gap in real-world clinical practice is multi-factorial, partially explained by technical and infrastructure hurdles, lack of IT resources, and no clear data-driven clinical utility analyses. Many of these barriers to adoption are being addressed with existing or emerging solutions [7,8,9,10]. Yet even with successful site specific model validation and successful integration/deployment in clinical workflows, a fundamental problem remains: what happens after the AI model goes into production?…”

Section: Introductionmentioning

confidence: 99%

CheXstray: Real-time Multi-Modal Data Concordance for Drift Detection in Medical Imaging AI

Soin¹,

Merkow²,

Jin³

et al. 2022

Preprint

View full text Add to dashboard Cite

Background/Motivation: Rapidly expanding Clinical AI applications worldwide have the potential to impact to all areas of medical practice. Medical imaging applications constitute a vast majority of approved clinical AI applications with more than 150 to date. Though healthcare systems are eager to adopt AI solutions after initial successes in model validation and deployment for clinical workflows, fundamental questions remain: Is the model still working as expected?, What is causing the change?, Is it time to intervene? While infrastructure and local validation barriers are addressed during implementation, the larger problem looms regarding how to monitor input data and track model performance in real-time production systems. Real-world healthcare data environments are dynamic which can disrupt AI model performance, referred to as "drift". For example, medical equipment updates, new imaging workflows, changes in patient and disease characteristics, and more, can change over time and cause drift. While there are dozens of tools for monitoring drift, all lack functionality for image-based deep learning model architectures, cannot use healthcare medical imaging data, and require contemporaneous ground truth which is rarely available in healthcare settings. As a result, healthcare systems are left without insight for ongoing medical imaging model performance which poses serious safety concerns for healthcare system leadership, government regulators, and patients.Methods: In this study, we utilize the CheXpert and PadChest open datasets to build, design, and test a medical imaging AI drift monitoring workflow that can track data and model drift in the absence of contemporaneous ground truth for a chest X-ray use case. Ten pathology labels were chosen and reconciled between the two datasets in order to fine-tune a densely connected convolutional neural network on PadChest after being pre-trained on frontal-only CheXpert images. The PadChest dataset and accompanying ground truth labels were chronologically arranged and split such that all finetuning and validation used only data between 2007-2013 while tests were performed on data between 2014-2017. To test the robustness of our approach, we simulate drift in a variety of experiments and observe measured performance and drift within a data stream. We measure performance using ground truth labels and compare our drift metric with actual model performance over time in the test dataset and identify correlative signals. Drift evaluation used DICOM metadata, image appearance, and model output scores as input. Variational autoencoders (VAE) were used to capture an appearance feature encoding by using the underlying latent space representation. For continuous features (e.g.

show abstract

Toward Generalizability in the Deployment of Artificial Intelligence in Radiology: Role of Computation Stress Testing to Overcome Underspecification

Cited by 82 publications

References 72 publications

Artificial intelligence and radiomics: fundamentals, applications, and challenges in immunotherapy

Artificial intelligence and radiomics: fundamentals, applications, and challenges in immunotherapy

Artificial Intelligence in Emergency Radiology: Where Are We Going?

CheXstray: Real-time Multi-Modal Data Concordance for Drift Detection in Medical Imaging AI

Contact Info

Product

Resources

About