Patterns of red blood cell utilization: Harnessing electronic health records data from the Information Standard for Blood and Transplant (ISBT) 128 system within the Biologics Effectiveness and Safety (BEST) initiative.

BACKGROUND: Current hemovigilance methods generally rely on survey data or administrative claims data utilizing billing and revenue codes, each of which has limitations. We used electronic health records (EHR) linked to blood bank data to comprehensively characterize red blood cell (RBC) utilization patterns and trends in three healthcare systems participating in the U.S. Food and Drug Administration Center for Biologics Evaluation and Research Biologics Effectiveness and Safety (BEST) initiative. METHODS: We used Information Standard for Blood and Transplant (ISBT) 128 codes linked to EHR from three healthcare systems data sources to identify and quantify RBC-transfused individuals, RBC transfusion episodes, transfused RBC units, and processing methods per year during 2012-2018. RESULTS: There were 577,822 RBC units transfused among 112,705 patients comprising 345,373 transfusion episodes between 2012 and 2018. Utilization in terms of RBC units and patients increased slightly in one and decreased slightly in the other two healthcare facilities. About 90% of RBC-transfused patients had 1 (~46%) or 2-5 (~42%)transfusion episodes in 2018. Among the small proportion of patients with ≥12 transfusion episodes per year, approximately 60% of episodes included only one RBC unit. All facilities used leukocyte-reduced RBCs during the study period whereas irradiated RBC utilization patterns differed across facilities. DISCUSSION: ISBT 128 codes and EHRs were used to observe patterns of RBC transfusion and modification methods at the unit level and patient level in three healthcare systems participating in the BEST initiative. This study shows that the ISBT 128 coding system in an EHR environment provides a feasible source for hemovigilance activities.

A DICOM Framework for Machine Learning and Processing Pipelines Against Real-time Radiology Images.

Real-time execution of machine learning (ML) pipelines on radiology images is difficult due to limited computing resources in clinical environments, whereas running them in research clusters requires efficient data transfer capabilities. We developed Niffler, an open-source Digital Imaging and Communications in Medicine (DICOM) framework that enables ML and processing pipelines in research clusters by efficiently retrieving images from the hospitals' PACS and extracting the metadata from the images. We deployed Niffler at our institution (Emory Healthcare, the largest healthcare network in the state of Georgia) and retrieved data from 715 scanners spanning 12 sites, up to 350 GB/day continuously in real-time as a DICOM data stream over the past 2 years. We also used Niffler to retrieve images bulk on-demand based on user-provided filters to facilitate several research projects. This paper presents the architecture and three such use cases of Niffler. First, we executed an IVC filter detection and segmentation pipeline on abdominal radiographs in real-time, which was able to classify 989 test images with an accuracy of 96.0%. Second, we applied the Niffler Metadata Extractor to understand the operational efficiency of individual MRI systems based on calculated metrics. We benchmarked the accuracy of the calculated exam time windows by comparing Niffler against the Clinical Data Warehouse (CDW). Niffler accurately identified the scanners' examination timeframes and idling times, whereas CDW falsely depicted several exam overlaps due to human errors. Third, with metadata extracted from the images by Niffler, we identified scanners with misconfigured time and reconfigured five scanners. Our evaluations highlight how Niffler enables real-time ML and processing pipelines in a research cluster.

Identification of Patients in Need of Advanced Care for Depression Using Data Extracted From a Statewide Health Information Exchange: A Machine Learning Approach.

BACKGROUND: As the most commonly occurring form of mental illness worldwide, depression poses significant health and economic burdens to both the individual and community. Different types of depression pose different levels of risk. Individuals who suffer from mild forms of depression may recover without any assistance or be effectively managed by primary care or family practitioners. However, other forms of depression are far more severe and require advanced care by certified mental health providers. However, identifying cases of depression that require advanced care may be challenging to primary care providers and health care team members whose skill sets run broad rather than deep. OBJECTIVE: This study aimed to leverage a comprehensive range of patient-level diagnostic, behavioral, and demographic data, as well as past visit history data from a statewide health information exchange to build decision models capable of predicting the need of advanced care for depression across patients presenting at Eskenazi Health, the public safety net health system for Marion County, Indianapolis, Indiana. METHODS: Patient-level diagnostic, behavioral, demographic, and past visit history data extracted from structured datasets were merged with outcome variables extracted from unstructured free-text datasets and were used to train random forest decision models that predicted the need of advanced care for depression across (1) the overall patient population and (2) various subsets of patients at higher risk for depression-related adverse events; patients with a past diagnosis of depression; patients with a Charlson comorbidity index of ≥1; patients with a Charlson comorbidity index of ≥2; and all unique patients identified across the 3 above-mentioned high-risk groups. RESULTS: The overall patient population consisted of 84,317 adult (aged ≥18 years) patients. A total of 6992 (8.29%) of these patients were in need of advanced care for depression. Decision models for high-risk patient groups yielded area under the curve (AUC) scores between 86.31% and 94.43%. The decision model for the overall patient population yielded a comparatively lower AUC score of 78.87%. The variance of optimal sensitivity and specificity for all decision models, as identified using Youden J Index, is as follows: sensitivity=68.79% to 83.91% and specificity=76.03% to 92.18%. CONCLUSIONS: This study demonstrates the ability to automate screening for patients in need of advanced care for depression across (1) an overall patient population or (2) various high-risk patient groups using structured datasets covering acute and chronic conditions, patient demographics, behaviors, and past visit history. Furthermore, these results show considerable potential to enable preventative care and can be easily integrated into existing clinical workflows to improve access to wraparound health care services.

A Platform for Innovation and Standards Evaluation: a Case Study from the OpenMRS Open-Source Radiology Information System.

Open-source development can provide a platform for innovation by seeking feedback from community members as well as providing tools and infrastructure to test new standards. Vendors of proprietary systems may delay adoption of new standards until there are sufficient incentives such as legal mandates or financial incentives to encourage/mandate adoption. Moreover, open-source systems in healthcare have been widely adopted in low- and middle-income countries and can be used to bridge gaps that exist in global health radiology. Since 2011, the authors, along with a community of open-source contributors, have worked on developing an open-source radiology information system (RIS) across two communities-OpenMRS and LibreHealth. The main purpose of the RIS is to implement core radiology workflows, on which others can build and test new radiology standards. This work has resulted in three major releases of the system, with current architectural changes driven by changing technology, development of new standards in health and imaging informatics, and changing user needs. At their core, both these communities are focused on building general-purpose EHR systems, but based on user contributions from the fringes, we have been able to create an innovative system that has been used by hospitals and clinics in four different countries. We provide an overview of the history of the LibreHealth RIS, the architecture of the system, overview of standards integration, describe challenges of developing an open-source product, and future directions. Our goal is to attract more participation and involvement to further develop the LibreHealth RIS into an Enterprise Imaging System that can be used in other clinical imaging including pathology and dermatology.

Datawiz-IN: Summer Research Experience for Health Data Science Training.

Background: Good science necessitates diverse perspectives to guide its progress. This study introduces Datawiz-IN, an educational initiative that fosters diversity and inclusion in AI skills training and research. Supported by a National Institutes of Health R25 grant from the National Library of Medicine, Datawiz-IN provided a comprehensive data science and machine learning research experience to students from underrepresented minority groups in medicine and computing. Methods: The program evaluation triangulated quantitative and qualitative data to measure representation, innovation, and experience. Diversity gains were quantified using demographic data analysis. Computational projects were systematically reviewed for research productivity. A mixed-methods survey gauged participant perspectives on skills gained, support quality, challenges faced, and overall sentiments. Results: The first cohort of 14 students in Summer 2023 demonstrated quantifiable increases in representation, with greater participation of women and minorities, evidencing the efficacy of proactive efforts to engage talent typically excluded from these fields. The student interns conducted innovative projects that elucidated disease mechanisms, enhanced clinical decision support systems, and analyzed health disparities. Conclusion: By illustrating how purposeful inclusion catalyzes innovation, Datawiz-IN offers a model for developing AI systems and research that reflect true diversity. Realizing the full societal benefits of AI requires sustaining pathways for historically excluded voices to help shape the field.

A Skills Assessment Pathways-Based Program Assessment Approach in Multidisciplinary Graduate Health Informatics.

Multidisciplinary graduate education programs are hard to assess because of interdependent competencies. Students in these programs come with diverse disciplinary undergraduate degrees, and it is critical to identify knowledge gaps among these diverse learner groups to provide support to fill these gaps. Health Informatics (HI) is a multidisciplinary field in which health, technology, and social science knowledge are foundational to building HI competencies. In 2017, the American Medical Informatics Association identified ten functional domains in which HI competencies are divided. Using pre/post-semester knowledge assessment surveys of graduate students (n=60) between August 2021 to May 2022 in one of the largest graduate HI programs in the United States, we identified courses (n=9) across the curriculum that help build HI-specific competencies. Using statistical analysis, we identified three skills pathways by correlating knowledge gained with course learning objectives and used this to modify the curriculum over four semesters. These skills pathways are connected through one or two courses, where students can choose electives or, in some instances, course modules or assignments that link the skills pathways. Moreover, there is a statistically significant difference in how students gain these skills depending on their prior training, even though they take the same set of courses. Gender and other demographics did not show statistical differences in skills gained. Additionally, we found that research assistantships and internships/practicums provide additional skills not covered in our HI curriculum. Our program assessment methodology and resulting curricular changes might be relevant to HI and other multidisciplinary graduate training programs.

AI Education for Fourth-Year Medical Students: Two-Year Experience of a Web-Based, Self-Guided Curriculum and Mixed Methods Study.

BACKGROUND: Artificial intelligence (AI) and machine learning (ML) are poised to have a substantial impact in the health care space. While a plethora of web-based resources exist to teach programming skills and ML model development, there are few introductory curricula specifically tailored to medical students without a background in data science or programming. Programs that do exist are often restricted to a specific specialty. OBJECTIVE: We hypothesized that a 1-month elective for fourth-year medical students, composed of high-quality existing web-based resources and a project-based structure, would empower students to learn about the impact of AI and ML in their chosen specialty and begin contributing to innovation in their field of interest. This study aims to evaluate the success of this elective in improving self-reported confidence scores in AI and ML. The authors also share our curriculum with other educators who may be interested in its adoption. METHODS: This elective was offered in 2 tracks: technical (for students who were already competent programmers) and nontechnical (with no technical prerequisites, focusing on building a conceptual understanding of AI and ML). Students established a conceptual foundation of knowledge using curated web-based resources and relevant research papers, and were then tasked with completing 3 projects in their chosen specialty: a data set analysis, a literature review, and an AI project proposal. The project-based nature of the elective was designed to be self-guided and flexible to each student's interest area and career goals. Students' success was measured by self-reported confidence in AI and ML skills in pre and postsurveys. Qualitative feedback on students' experiences was also collected. RESULTS: This web-based, self-directed elective was offered on a pass-or-fail basis each month to fourth-year students at Emory University School of Medicine beginning in May 2021. As of June 2022, a total of 19 students had successfully completed the elective, representing a wide range of chosen specialties: diagnostic radiology (n=3), general surgery (n=1), internal medicine (n=5), neurology (n=2), obstetrics and gynecology (n=1), ophthalmology (n=1), orthopedic surgery (n=1), otolaryngology (n=2), pathology (n=2), and pediatrics (n=1). Students' self-reported confidence scores for AI and ML rose by 66% after this 1-month elective. In qualitative surveys, students overwhelmingly reported enthusiasm and satisfaction with the course and commented that the self-direction and flexibility and the project-based design of the course were essential. CONCLUSIONS: Course participants were successful in diving deep into applications of AI in their widely-ranging specialties, produced substantial project deliverables, and generally reported satisfaction with their elective experience. The authors are hopeful that a brief, 1-month investment in AI and ML education during medical school will empower this next generation of physicians to pave the way for AI and ML innovation in health care.

Theory of radiologist interaction with instant messaging decision support tools: A sequential-explanatory study.

Radiology specific clinical decision support systems (CDSS) and artificial intelligence are poorly integrated into the radiologist workflow. Current research and development efforts of radiology CDSS focus on 4 main interventions, based around exam centric time points-after image acquisition, intra-report support, post-report analysis, and radiology workflow adjacent. We review the literature surrounding CDSS tools in these time points, requirements for CDSS workflow augmentation, and technologies that support clinician to computer workflow augmentation. We develop a theory of radiologist-decision tool interaction using a sequential explanatory study design. The study consists of 2 phases, the first a quantitative survey and the second a qualitative interview study. The phase 1 survey identifies differences between average users and radiologist users in software interventions using the User Acceptance of Information Technology: Toward a Unified View (UTAUT) framework. Phase 2 semi-structured interviews provide narratives on why these differences are found. To build this theory, we propose a novel solution called Radibot-a conversational agent capable of engaging clinicians with CDSS as an assistant using existing instant messaging systems supporting hospital communications. This work contributes an understanding of how radiologist-users differ from the average user and can be utilized by software developers to increase satisfaction of CDSS tools within radiology.

Identifying and improving the "ground truth" of race in disparities research through improved EMR data reporting. A systematic review.

BACKGROUND: Studies about racial disparities in healthcare are increasing in quantity; however, they are subject to vast differences in definition, classification, and utilization of race/ethnicity data. Improved standardization of this information can strengthen conclusions drawn from studies using such data. The objective of this study is to examine how data related to race/ethnicity are recorded in research through examining articles on race/ethnicity health disparities and examine problems and solutions in data reporting that may impact overall data quality. METHODS: In this systematic review, Business Source Complete, Embase.com, IEEE Xplore, PubMed, Scopus and Web of Science Core Collection were searched for relevant articles published from 2000 to 2020. Search terms related to the concepts of electronic medical records, race/ethnicity, and data entry related to race/ethnicity were used. Exclusion criteria included articles not in the English language and those describing pediatric populations. Data were extracted from published articles. This review was organized and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement for systematic reviews. FINDINGS: In this systematic review, 109 full text articles were reviewed. Weaknesses and possible solutions have been discussed in current literature, with the predominant problem and solution as follows: the electronic medical record (EMR) is vulnerable to inaccuracies and incompleteness in the methods that research staff collect this data; however, improved standardization of the collection and use of race data in patient care may help alleviate these inaccuracies. INTERPRETATION: Conclusions drawn from large datasets concerning peoples of certain race/ethnic groups should be made cautiously, and a careful review of the methodology of each publication should be considered prior to implementation in patient care.

Synthetically enhanced: unveiling synthetic data's potential in medical imaging research.

BACKGROUND: Chest X-rays (CXR) are essential for diagnosing a variety of conditions, but when used on new populations, model generalizability issues limit their efficacy. Generative AI, particularly denoising diffusion probabilistic models (DDPMs), offers a promising approach to generating synthetic images, enhancing dataset diversity. This study investigates the impact of synthetic data supplementation on the performance and generalizability of medical imaging research. METHODS: The study employed DDPMs to create synthetic CXRs conditioned on demographic and pathological characteristics from the CheXpert dataset. These synthetic images were used to supplement training datasets for pathology classifiers, with the aim of improving their performance. The evaluation involved three datasets (CheXpert, MIMIC-CXR, and Emory Chest X-ray) and various experiments, including supplementing real data with synthetic data, training with purely synthetic data, and mixing synthetic data with external datasets. Performance was assessed using the area under the receiver operating curve (AUROC). FINDINGS: Adding synthetic data to real datasets resulted in a notable increase in AUROC values (up to 0.02 in internal and external test sets with 1000% supplementation, p-value <0.01 in all instances). When classifiers were trained exclusively on synthetic data, they achieved performance levels comparable to those trained on real data with 200%-300% data supplementation. The combination of real and synthetic data from different sources demonstrated enhanced model generalizability, increasing model AUROC from 0.76 to 0.80 on the internal test set (p-value <0.01). INTERPRETATION: Synthetic data supplementation significantly improves the performance and generalizability of pathology classifiers in medical imaging. FUNDING: Dr. Gichoya is a 2022 Robert Wood Johnson Foundation Harold Amos Medical Faculty Development Program and declares support from RSNA Health Disparities grant (#EIHD2204), Lacuna Fund (#67), Gordon and Betty Moore Foundation, NIH (NIBIB) MIDRC grant under contracts 75N92020C00008 and 75N92020C00021, and NHLBI Award Number R01HL167811.

A multimodal framework for extraction and fusion of satellite images and public health data.

In low- and middle-income countries, the substantial costs associated with traditional data collection pose an obstacle to facilitating decision-making in the field of public health. Satellite imagery offers a potential solution, but the image extraction and analysis can be costly and requires specialized expertise. We introduce SatelliteBench, a scalable framework for satellite image extraction and vector embeddings generation. We also propose a novel multimodal fusion pipeline that utilizes a series of satellite imagery and metadata. The framework was evaluated generating a dataset with a collection of 12,636 images and embeddings accompanied by comprehensive metadata, from 81 municipalities in Colombia between 2016 and 2018. The dataset was then evaluated in 3 tasks: including dengue case prediction, poverty assessment, and access to education. The performance showcases the versatility and practicality of SatelliteBench, offering a reproducible, accessible and open tool to enhance decision-making in public health.

MedShift: Automated Identification of Shift Data for Medical Image Dataset Curation.

Automated curation of noisy external data in the medical domain has long been in high demand, as AI technologies need to be validated using various sources with clean, annotated data. Identifying the variance between internal and external sources is a fundamental step in curating a high-quality dataset, as the data distributions from different sources can vary significantly and subsequently affect the performance of AI models. The primary challenges for detecting data shifts are - (1) accessing private data across healthcare institutions for manual detection and (2) the lack of automated approaches to learn efficient shift-data representation without training samples. To overcome these problems, we propose an automated pipeline called MedShift to detect top-level shift samples and evaluate the significance of shift data without sharing data between internal and external organizations. MedShift employs unsupervised anomaly detectors to learn the internal distribution and identify samples showing significant shiftness for external datasets, and then compares their performance. To quantify the effects of detected shift data, we train a multi-class classifier that learns internal domain knowledge and evaluates the classification performance for each class in external domains after dropping the shift data. We also propose a data quality metric to quantify the dissimilarity between internal and external datasets. We verify the efficacy of MedShift using musculoskeletal radiographs (MURA) and chest X-ray datasets from multiple external sources. Our experiments show that our proposed shift data detection pipeline can be beneficial for medical centers to curate high-quality datasets more efficiently.

Evaluation of accessibility of open-source EHRs for visually impaired users.

This study investigates the accessibility of open-source electronic health record (EHR) systems for individuals who are visually impaired or blind. Ensuring the accessibility of EHRs to visually impaired users is critical for the diversity, equity, and inclusion of all users. The study used a combination of automated and manual accessibility testing with screen readers to evaluate the accessibility of three widely used open-source EHR systems. We used three popular screen readers - JAWS (Windows), NVDA (Windows), and Apple VoiceOver (OSX) to evaluate accessibility. The evaluation revealed that although each of the three EHR systems was partially accessible, there is room for improvement, particularly regarding keyboard navigation and screen reader compatibility. The study concludes with recommendations for making EHR systems more inclusive for all users and more accessible.

Foundational domains and competencies for baccalaureate health informatics education.

BACKGROUND: Foundational domains are the building blocks of educational programs. The lack of foundational domains in undergraduate health informatics (HI) education can adversely affect the development of rigorous curricula and may impede the attainment of CAHIIM accreditation of academic programs. OBJECTIVE: This White Paper presents foundational domains developed by AMIA's Academic Forum Baccalaureate Education Committee (BEC) which include corresponding competencies (knowledge, skills, and attitudes) that are intended for curriculum development and CAHIIM accreditation quality assessment for undergraduate education in applied health informatics. METHODS: The AMIA BEC used the previously published master's foundational domains as a guide to creating a set of competencies for health informatics at the undergraduate level to assess graduates from undergraduate health informatics programs for competence at graduation. A consensus method was used to adapt the domains for undergraduate level course work and harmonize the foundational domains with the currently adapted domains for HI master's education. RESULTS: Ten foundational domains were developed to support the development and evaluation of baccalaureate health informatics education. DISCUSSION: This article will inform future work towards building CAHIIM accreditation standards to ensure that higher education institutions meet acceptable levels of quality for undergraduate health informatics education.

Generalizability of Human Activity Recognition Machine Learning Models from non-Parkinson's to Parkinson's Disease Patients.

Recent evidence shows that high-intensity exercises reduce tremors and stiffness in Parkinson's disease (PD). However, there is insufficient evidence on the types of exercises; in effect, high-intensity may be a personalized measure. Recent progress in automated Human Activity Recognition using machine learning (ML) models shows potential for better monitoring of PD patients. However, ML models must be calibrated to ignore tremors and accurately identify activity and its intensity. We report findings from a study where we trained ML models using data from medically validated triple synchronous sensors connected to 8 non-PD subjects performing 32 exercises. We then tested the models to identify exercises performed by 8 PD patients at different stages of the disease. Our analysis shows that better data preprocessing before modeling can provide some model generalizability. However, it is extremely challenging, as the models work with high accuracy on one group (Healthy or PD patients) (F1=0.88-0.94) but not on both groups.Clinical relevance-Patients with Parkinson's and other motor-generative diseases can now accurately measure physical activity with machine learning approaches. Clinicians, caregivers, and apps can make accurate, personalized exercise recommendations to augment medications that reduce tremors and stiffness.

The NeoRoo mobile app: Initial design and prototyping of an Android-based digital health tool to support Kangaroo Mother Care in low/middle-income countries (LMICs).

Premature birth and neonatal mortality are significant global health challenges, with 15 million premature births annually and an estimated 2.5 million neonatal deaths. Approximately 90% of preterm births occur in low/middle income countries, particularly within the global regions of sub-Saharan Africa and South Asia. Neonatal hypothermia is a common and significant cause of morbidity and mortality among premature and low birth weight infants, particularly in low/middle-income countries where rates of premature delivery are high, and access to health workers, medical commodities, and other resources is limited. Kangaroo Mother Care/Skin-to-Skin care has been shown to significantly reduce the incidence of neonatal hypothermia and improve survival rates among premature infants, but there are significant barriers to its implementation, especially in low/middle-income countries (LMICs). The paper proposes the use of a multidisciplinary approach to develop an integrated mHealth solution to overcome the barriers and challenges to the implementation of Kangaroo Mother Care/Skin-to-skin care (KMC/STS) in LMICs. The innovation is an integrated mHealth platform that features a wearable biomedical device (NeoWarm) and an Android-based mobile application (NeoRoo) with customized user interfaces that are targeted specifically to parents/family stakeholders and healthcare providers, respectively. This publication describes the iterative, human-centered design and participatory development of a high-fidelity prototype of the NeoRoo mobile application. The aim of this study was to design and develop an initial ("A") version of the Android-based NeoRoo mobile app specifically to support the use case of KMC/STS in health facilities in Kenya. Key functions and features are highlighted. The proposed solution leverages the promise of digital health to overcome identified barriers and challenges to the implementation of KMC/STS in LMICs and aims to equip parents and healthcare providers of prematurely born infants with the tools and resources needed to improve the care provided to premature and low birthweight babies. It is hoped that, when implemented and scaled as part of a thoughtful, strategic, cross-disciplinary approach to reduction of global rates of neonatal mortality, NeoRoo will prove to be a useful tool within the toolkit of parents, health workers, and program implementors.

"Shortcuts" Causing Bias in Radiology Artificial Intelligence: Causes, Evaluation, and Mitigation.

Despite the expert-level performance of artificial intelligence (AI) models for various medical imaging tasks, real-world performance failures with disparate outputs for various subgroups limit the usefulness of AI in improving patients' lives. Many definitions of fairness have been proposed, with discussions of various tensions that arise in the choice of an appropriate metric to use to evaluate bias; for example, should one aim for individual or group fairness? One central observation is that AI models apply "shortcut learning" whereby spurious features (such as chest tubes and portable radiographic markers on intensive care unit chest radiography) on medical images are used for prediction instead of identifying true pathology. Moreover, AI has been shown to have a remarkable ability to detect protected attributes of age, sex, and race, while the same models demonstrate bias against historically underserved subgroups of age, sex, and race in disease diagnosis. Therefore, an AI model may take shortcut predictions from these correlations and subsequently generate an outcome that is biased toward certain subgroups even when protected attributes are not explicitly used as inputs into the model. As a result, these subgroups became nonprivileged subgroups. In this review, the authors discuss the various types of bias from shortcut learning that may occur at different phases of AI model development, including data bias, modeling bias, and inference bias. The authors thereafter summarize various tool kits that can be used to evaluate and mitigate bias and note that these have largely been applied to nonmedical domains and require more evaluation for medical AI. The authors then summarize current techniques for mitigating bias from preprocessing (data-centric solutions) and during model development (computational solutions) and postprocessing (recalibration of learning). Ongoing legal changes where the use of a biased model will be penalized highlight the necessity of understanding, detecting, and mitigating biases from shortcut learning and will require diverse research teams looking at the whole AI pipeline.

AI pitfalls and what not to do: mitigating bias in AI.

Various forms of artificial intelligence (AI) applications are being deployed and used in many healthcare systems. As the use of these applications increases, we are learning the failures of these models and how they can perpetuate bias. With these new lessons, we need to prioritize bias evaluation and mitigation for radiology applications; all the while not ignoring the impact of changes in the larger enterprise AI deployment which may have downstream impact on performance of AI models. In this paper, we provide an updated review of known pitfalls causing AI bias and discuss strategies for mitigating these biases within the context of AI deployment in the larger healthcare enterprise. We describe these pitfalls by framing them in the larger AI lifecycle from problem definition, data set selection and curation, model training and deployment emphasizing that bias exists across a spectrum and is a sequela of a combination of both human and machine factors.

Ability of artificial intelligence to identify self-reported race in chest x-ray using pixel intensity counts.

Purpose: Prior studies show convolutional neural networks predicting self-reported race using x-rays of chest, hand and spine, chest computed tomography, and mammogram. We seek an understanding of the mechanism that reveals race within x-ray images, investigating the possibility that race is not predicted using the physical structure in x-ray images but is embedded in the grayscale pixel intensities. Approach: Retrospective full year 2021, 298,827 AP/PA chest x-ray images from 3 academic health centers across the United States and MIMIC-CXR, labeled by self-reported race, were used in this study. The image structure is removed by summing the number of each grayscale value and scaling to percent per image (PPI). The resulting data are tested using multivariate analysis of variance (MANOVA) with Bonferroni multiple-comparison adjustment and class-balanced MANOVA. Machine learning (ML) feed-forward networks (FFN) and decision trees were built to predict race (binary Black or White and binary Black or other) using only grayscale value counts. Stratified analysis by body mass index, age, sex, gender, patient type, make/model of scanner, exposure, and kilovoltage peak setting was run to study the impact of these factors on race prediction following the same methodology. Results: MANOVA rejects the null hypothesis that classes are the same with 95% confidence (F 7.38, P<0.0001) and balanced MANOVA (F 2.02, P<0.0001). The best FFN performance is limited [area under the receiver operating characteristic (AUROC) of 69.18%]. Gradient boosted trees predict self-reported race using grayscale PPI (AUROC 77.24%). Conclusions: Within chest x-rays, pixel intensity value counts alone are statistically significant indicators and enough for ML classification tasks of patient self-reported race.

Margin-aware intraclass novelty identification for medical images.

Purpose: Existing anomaly detection methods focus on detecting interclass variations while medical image novelty identification is more challenging in the presence of intraclass variations. For example, a model trained with normal chest x-ray and common lung abnormalities is expected to discover and flag idiopathic pulmonary fibrosis, which is a rare lung disease and unseen during training. The nuances of intraclass variations and lack of relevant training data in medical image analysis pose great challenges for existing anomaly detection methods. Approach: We address the above challenges by proposing a hybrid model-transformation-based embedding learning for novelty detection (TEND), which combines the merits of classifier-based approach and AutoEncoder (AE)-based approach. Training TEND consists of two stages. In the first stage, we learn in-distribution embeddings with an AE via the unsupervised reconstruction. In the second stage, we learn a discriminative classifier to distinguish in-distribution data and the transformed counterparts. Additionally, we propose a margin-aware objective to pull in-distribution data in a hypersphere while pushing away the transformed data. Eventually, the weighted sum of class probability and the distance to margin constitutes the anomaly score. Results: Extensive experiments are performed on three public medical image datasets with the one-vs-rest setup (namely one class as in-distribution data and the left as intraclass out-of-distribution data) and the rest-vs-one setup. Additional experiments on generated intraclass out-of-distribution data with unused transformations are implemented on the datasets. The quantitative results show competitive performance as compared to the state-of-the-art approaches. Provided qualitative examples further demonstrate the effectiveness of TEND. Conclusion: Our anomaly detection model TEND can effectively identify the challenging intraclass out-of-distribution medical images in an unsupervised fashion. It can be applied to discover unseen medical image classes and serve as the abnormal data screening for downstream medical tasks. The corresponding code is available at https://github.com/XiaoyuanGuo/TEND_MedicalNoveltyDetection.