Addressing fairness in artificial intelligence for medical imaging

Generating multi-pathological and multi-modal images and labels for brain MRI

The last few years have seen a boom in using generative models to augment real datasets, as synthetic data can effectively model real data distributions and provide privacy-preserving, shareable datasets that can be used to train deep learning models. However, most of these methods are 2D and provide synthetic datasets that come, at most, with categorical annotations. The generation of paired images and segmentation samples that can be used in downstream, supervised segmentation tasks remains fairly uncharted territory. This work proposes a two-stage generative model capable of producing 2D and 3D semantic label maps and corresponding multi-modal images. We use a latent diffusion model for label synthesis and a VAE-GAN for semantic image synthesis. Synthetic datasets provided by this model are shown to work in a wide variety of segmentation tasks, supporting small, real datasets or fully replacing them while maintaining good performance. We also demonstrate its ability to improve downstream performance on out-of-distribution data.

Enhancing neuro-oncology care through equity-driven applications of artificial intelligence

The disease course and clinical outcome for brain tumor patients depend not only on the molecular and histological features of the tumor but also on the patient's demographics and social determinants of health. While current investigations in neuro-oncology have broadly utilized artificial intelligence (AI) to enrich tumor diagnosis and more accurately predict treatment response, postoperative complications, and survival, equity-driven applications of AI have been limited. However, AI applications to advance health equity in the broader medical field have the potential to serve as practical blueprints to address known disparities in neuro-oncologic care. In this consensus review, we will describe current applications of AI in neuro-oncology, postulate viable AI solutions for the most pressing inequities in neuro-oncology based on broader literature, propose a framework for the effective integration of equity into AI-based neuro-oncology research, and close with the limitations of AI.

Deep learning segmentation of mandible with lower dentition from cone beam CT

OBJECTIVES

This study aimed to train a 3D U-Net convolutional neural network (CNN) for mandible and lower dentition segmentation from cone-beam computed tomography (CBCT) scans.

METHODS

In an ambispective cross-sectional design, CBCT scans from two hospitals (2009-2019 and 2021-2022) constituted an internal dataset and external validation set, respectively. Manual segmentation informed CNN training, and evaluations employed Dice similarity coefficient (DSC) for volumetric accuracy. A blinded oral maxillofacial surgeon performed qualitative grading of CBCT scans and object meshes. Statistical analyses included independent t-tests and ANOVA tests to compare DSC across patient subgroups of gender, race, body mass index (BMI), test dataset used, age, and degree of metal artifact. Tests were powered for a minimum detectable difference in DSC of 0.025, with alpha of 0.05 and power level of 0.8.

RESULTS

648 CBCT scans from 490 patients were included in the study. The CNN achieved high accuracy (average DSC: 0.945 internal, 0.940 external). No DSC differences were observed between test set used, gender, BMI, and race. Significant differences in DSC were identified based on age group and the degree of metal artifact. The majority (80%) of object meshes produced by both manual and automatic segmentation were rated as acceptable or higher quality.

CONCLUSION

We developed a model for automatic mandible and lower dentition segmentation from CBCT scans in a demographically diverse cohort including a high degree of metal artifacts. The model demonstrated good accuracy on internal and external test sets, with majority acceptable quality from a clinical grader.

Biomedical Data Science, Artificial Intelligence, and Ethics: Navigating Challenges in the Face of Explosive Growth

Advances in biomedical data science and artificial intelligence (AI) are profoundly changing the landscape of healthcare. This article reviews the ethical issues that arise with the development of AI technologies, including threats to privacy, data security, consent, and justice, as they relate to donors of tissue and data. It also considers broader societal obligations, including the importance of assessing the unintended consequences of AI research in biomedicine. In addition, this article highlights the challenge of rapid AI development against the backdrop of disparate regulatory frameworks, calling for a global approach to address concerns around data misuse, unintended surveillance, and the equitable distribution of AI's benefits and burdens. Finally, a number of potential solutions to these ethical quandaries are offered. Namely, the merits of advocating for a collaborative, informed, and flexible regulatory approach that balances innovation with individual rights and public welfare, fostering a trustworthy AI-driven healthcare ecosystem, are discussed.

Fair AI-powered orthopedic image segmentation: addressing bias and promoting equitable healthcare

AI-powered segmentation of hip and knee bony anatomy has revolutionized orthopedics, transforming pre-operative planning and post-operative assessment. Despite the remarkable advancements in AI algorithms for medical imaging, the potential for biases inherent within these models remains largely unexplored. This study tackles these concerns by thoroughly re-examining AI-driven segmentation for hip and knee bony anatomy. While advanced imaging modalities like CT and MRI offer comprehensive views, plain radiographs (X-rays) predominate the standard initial clinical assessment due to their widespread availability, low cost, and rapid acquisition. Hence, we focused on plain radiographs to ensure the utilization of our contribution in diverse healthcare settings, including those with limited access to advanced imaging technologies. This work provides insights into the underlying causes of biases in AI-based knee and hip image segmentation through an extensive evaluation, presenting targeted mitigation strategies to alleviate biases related to sex, race, and age, using an automatic segmentation that is fair, impartial, and safe in the context of AI. Our contribution can enhance inclusivity, ethical practices, equity, and an unbiased healthcare environment with advanced clinical outcomes, aiding decision-making and osteoarthritis research. Furthermore, we have made all the codes and datasets publicly and freely accessible to promote open scientific research.

Unmasking bias in artificial intelligence: a systematic review of bias detection and mitigation strategies in electronic health record-based models

Objectives

Leveraging artificial intelligence (AI) in conjunction with electronic health records (EHRs) holds transformative potential to improve healthcare. However, addressing bias in AI, which risks worsening healthcare disparities, cannot be overlooked. This study reviews methods to handle various biases in AI models developed using EHR data.

Materials and Methods

We conducted a systematic review following the Preferred Reporting Items for Systematic Reviews and Meta-analyses guidelines, analyzing articles from PubMed, Web of Science, and IEEE published between January 01, 2010 and December 17, 2023. The review identified key biases, outlined strategies for detecting and mitigating bias throughout the AI model development, and analyzed metrics for bias assessment.

Results

Of the 450 articles retrieved, 20 met our criteria, revealing 6 major bias types: algorithmic, confounding, implicit, measurement, selection, and temporal. The AI models were primarily developed for predictive tasks, yet none have been deployed in real-world healthcare settings. Five studies concentrated on the detection of implicit and algorithmic biases employing fairness metrics like statistical parity, equal opportunity, and predictive equity. Fifteen studies proposed strategies for mitigating biases, especially targeting implicit and selection biases. These strategies, evaluated through both performance and fairness metrics, predominantly involved data collection and preprocessing techniques like resampling and reweighting.

Discussion

This review highlights evolving strategies to mitigate bias in EHR-based AI models, emphasizing the urgent need for both standardized and detailed reporting of the methodologies and systematic real-world testing and evaluation. Such measures are essential for gauging models' practical impact and fostering ethical AI that ensures fairness and equity in healthcare.

Recognition of Patient Gender: A Machine Learning Preliminary Analysis Using Heart Sounds from Children and Adolescents

Research has shown that X-rays and fundus images can classify gender, age group, and race, raising concerns about bias and fairness in medical AI applications. However, the potential for physiological sounds to classify sociodemographic traits has not been investigated. Exploring this gap is crucial for understanding the implications and ensuring fairness in the field of medical sound analysis. We aimed to develop classifiers to determine gender (men/women) based on heart sound recordings and using machine learning (ML). Data-driven ML analysis. We utilized the open-access CirCor DigiScope Phonocardiogram Dataset obtained from cardiac screening programs in Brazil. Volunteers < 21 years of age. Each participant completed a questionnaire and underwent a clinical examination, including electronic auscultation at four cardiac points: aortic (AV), mitral (MV), pulmonary (PV), and tricuspid (TV). We used Mel-frequency cepstral coefficients (MFCCs) to develop the ML classifiers. From each patient and from each auscultation sound recording, we extracted 10 MFCCs. In sensitivity analysis, we additionally extracted 20, 30, 40, and 50 MFCCs. The most effective gender classifier was developed using PV recordings (AUC ROC = 70.3%). The second best came from MV recordings (AUC ROC = 58.8%). AV and TV recordings produced classifiers with an AUC ROC of 56.4% and 56.1%, respectively. Using more MFCCs did not substantially improve the classifiers. It is possible to classify between males and females using phonocardiogram data. As health-related audio recordings become more prominent in ML applications, research is required to explore if these recordings contain signals that could distinguish sociodemographic features.

From data to diagnosis: skin cancer image datasets for artificial intelligence

Artificial intelligence (AI) solutions for skin cancer diagnosis continue to gain momentum, edging closer towards broad clinical use. These AI models, particularly deep-learning architectures, require large digital image datasets for development. This review provides an overview of the datasets used to develop AI algorithms and highlights the importance of dataset transparency for the evaluation of algorithm generalizability across varying populations and settings. Current challenges for curation of clinically valuable datasets are detailed, which include dataset shifts arising from demographic variations and differences in data collection methodologies, along with inconsistencies in labelling. These shifts can lead to differential algorithm performance, compromise of clinical utility, and the propagation of discriminatory biases when developed algorithms are implemented in mismatched populations. Limited representation of rare skin cancers and minoritized groups in existing datasets are highlighted, which can further skew algorithm performance. Strategies to address these challenges are presented, which include improving transparency, representation and interoperability. Federated learning and generative methods, which may improve dataset size and diversity without compromising privacy, are also examined. Lastly, we discuss model-level techniques that may address biases entrained through the use of datasets derived from routine clinical care. As the role of AI in skin cancer diagnosis becomes more prominent, ensuring the robustness of underlying datasets is increasingly important.

Fairer AI in ophthalmology via implicit fairness learning for mitigating sexism and ageism

The transformative role of artificial intelligence (AI) in various fields highlights the need for it to be both accurate and fair. Biased medical AI systems pose significant potential risks to achieving fair and equitable healthcare. Here, we show an implicit fairness learning approach to build a fairer ophthalmology AI (called FairerOPTH) that mitigates sex (biological attribute) and age biases in AI diagnosis of eye diseases. Specifically, FairerOPTH incorporates the causal relationship between fundus features and eye diseases, which is relatively independent of sensitive attributes such as race, sex, and age. We demonstrate on a large and diverse collected dataset that FairerOPTH significantly outperforms several state-of-the-art approaches in terms of diagnostic accuracy and fairness for 38 eye diseases in ultra-widefield imaging and 16 eye diseases in narrow-angle imaging. This work demonstrates the significant potential of implicit fairness learning in promoting equitable treatment for patients regardless of their sex or age.

A survey of recent methods for addressing AI fairness and bias in biomedicine

OBJECTIVES

Artificial intelligence (AI) systems have the potential to revolutionize clinical practices, including improving diagnostic accuracy and surgical decision-making, while also reducing costs and manpower. However, it is important to recognize that these systems may perpetuate social inequities or demonstrate biases, such as those based on race or gender. Such biases can occur before, during, or after the development of AI models, making it critical to understand and address potential biases to enable the accurate and reliable application of AI models in clinical settings. To mitigate bias concerns during model development, we surveyed recent publications on different debiasing methods in the fields of biomedical natural language processing (NLP) or computer vision (CV). Then we discussed the methods, such as data perturbation and adversarial learning, that have been applied in the biomedical domain to address bias.

METHODS

We performed our literature search on PubMed, ACM digital library, and IEEE Xplore of relevant articles published between January 2018 and December 2023 using multiple combinations of keywords. We then filtered the result of 10,041 articles automatically with loose constraints, and manually inspected the abstracts of the remaining 890 articles to identify the 55 articles included in this review. Additional articles in the references are also included in this review. We discuss each method and compare its strengths and weaknesses. Finally, we review other potential methods from the general domain that could be applied to biomedicine to address bias and improve fairness.

RESULTS

The bias of AIs in biomedicine can originate from multiple sources such as insufficient data, sampling bias and the use of health-irrelevant features or race-adjusted algorithms. Existing debiasing methods that focus on algorithms can be categorized into distributional or algorithmic. Distributional methods include data augmentation, data perturbation, data reweighting methods, and federated learning. Algorithmic approaches include unsupervised representation learning, adversarial learning, disentangled representation learning, loss-based methods and causality-based methods.

Perfect Match: Radiomics and Artificial Intelligence in Cardiac Imaging

Cardiovascular diseases remain a significant health burden, with imaging modalities like echocardiography, cardiac computed tomography, and cardiac magnetic resonance imaging playing a crucial role in diagnosis and prognosis. However, the inherent heterogeneity of these diseases poses challenges, necessitating advanced analytical methods like radiomics and artificial intelligence. Radiomics extracts quantitative features from medical images, capturing intricate patterns and subtle variations that may elude visual inspection. Artificial intelligence techniques, including deep learning, can analyze these features to generate knowledge, define novel imaging biomarkers, and support diagnostic decision-making and outcome prediction. Radiomics and artificial intelligence thus hold promise for significantly enhancing diagnostic and prognostic capabilities in cardiac imaging, paving the way for more personalized and effective patient care. This review explores the synergies between radiomics and artificial intelligence in cardiac imaging, following the radiomics workflow and introducing concepts from both domains. Potential clinical applications, challenges, and limitations are discussed, along with solutions to overcome them.

CheXmask: a large-scale dataset of anatomical segmentation masks for multi-center chest x-ray images

The development of successful artificial intelligence models for chest X-ray analysis relies on large, diverse datasets with high-quality annotations. While several databases of chest X-ray images have been released, most include disease diagnosis labels but lack detailed pixel-level anatomical segmentation labels. To address this gap, we introduce an extensive chest X-ray multi-center segmentation dataset with uniform and fine-grain anatomical annotations for images coming from five well-known publicly available databases: ChestX-ray8, CheXpert, MIMIC-CXR-JPG, Padchest, and VinDr-CXR, resulting in 657,566 segmentation masks. Our methodology utilizes the HybridGNet model to ensure consistent and high-quality segmentations across all datasets. Rigorous validation, including expert physician evaluation and automatic quality control, was conducted to validate the resulting masks. Additionally, we provide individualized quality indices per mask and an overall quality estimation per dataset. This dataset serves as a valuable resource for the broader scientific community, streamlining the development and assessment of innovative methodologies in chest X-ray analysis.

Application of Artificial Intelligence Methods on Osteoporosis Classification with Radiographs-A Systematic Review

Osteoporosis is a complex endocrine disease characterized by a decline in bone mass and microstructural integrity. It constitutes a major global health problem. Recent progress in the field of artificial intelligence (AI) has opened new avenues for the effective diagnosis of osteoporosis via radiographs. This review investigates the application of AI classification of osteoporosis in radiographs. A comprehensive exploration of electronic repositories (ClinicalTrials.gov, Web of Science, PubMed, MEDLINE) was carried out in adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 statement (PRISMA). A collection of 31 articles was extracted from these repositories and their significant outcomes were consolidated and outlined. This encompassed insights into anatomical regions, the specific machine learning methods employed, the effectiveness in predicting BMD, and categorizing osteoporosis. Through analyzing the respective studies, we evaluated the effectiveness and limitations of AI osteoporosis classification in radiographs. The pooled reported accuracy, sensitivity, and specificity of osteoporosis classification ranges from 66.1% to 97.9%, 67.4% to 100.0%, and 60.0% to 97.5% respectively. This review underscores the potential of AI osteoporosis classification and offers valuable insights for future research endeavors, which should focus on addressing the challenges in technical and clinical integration to facilitate practical implementation of this technology.

Retinal imaging for the assessment of stroke risk: a systematic review

BACKGROUND

Stroke is a leading cause of morbidity and mortality. Retinal imaging allows non-invasive assessment of the microvasculature. Consequently, retinal imaging is a technology which is garnering increasing attention as a means of assessing cardiovascular health and stroke risk.

METHODS

A biomedical literature search was performed to identify prospective studies that assess the role of retinal imaging derived biomarkers as indicators of stroke risk.

RESULTS

Twenty-four studies were included in this systematic review. The available evidence suggests that wider retinal venules, lower fractal dimension, increased arteriolar tortuosity, presence of retinopathy, and presence of retinal emboli are associated with increased likelihood of stroke. There is weaker evidence to suggest that narrower arterioles and the presence of individual retinopathy traits such as microaneurysms and arteriovenous nicking indicate increased stroke risk. Our review identified three models utilizing artificial intelligence algorithms for the analysis of retinal images to predict stroke. Two of these focused on fundus photographs, whilst one also utilized optical coherence tomography (OCT) technology images. The constructed models performed similarly to conventional risk scores but did not significantly exceed their performance. Only two studies identified in this review used OCT imaging, despite the higher dimensionality of this data.

CONCLUSION

Whilst there is strong evidence that retinal imaging features can be used to indicate stroke risk, there is currently no predictive model which significantly outperforms conventional risk scores. To develop clinically useful tools, future research should focus on utilization of deep learning algorithms, validation in external cohorts, and analysis of OCT images.

Unmasking bias in artificial intelligence: a systematic review of bias detection and mitigation strategies in electronic health record-based models

OBJECTIVES

Leveraging artificial intelligence (AI) in conjunction with electronic health records (EHRs) holds transformative potential to improve healthcare. However, addressing bias in AI, which risks worsening healthcare disparities, cannot be overlooked. This study reviews methods to handle various biases in AI models developed using EHR data.

MATERIALS AND METHODS

We conducted a systematic review following the Preferred Reporting Items for Systematic Reviews and Meta-analyses guidelines, analyzing articles from PubMed, Web of Science, and IEEE published between January 01, 2010 and December 17, 2023. The review identified key biases, outlined strategies for detecting and mitigating bias throughout the AI model development, and analyzed metrics for bias assessment.

RESULTS

Of the 450 articles retrieved, 20 met our criteria, revealing 6 major bias types: algorithmic, confounding, implicit, measurement, selection, and temporal. The AI models were primarily developed for predictive tasks, yet none have been deployed in real-world healthcare settings. Five studies concentrated on the detection of implicit and algorithmic biases employing fairness metrics like statistical parity, equal opportunity, and predictive equity. Fifteen studies proposed strategies for mitigating biases, especially targeting implicit and selection biases. These strategies, evaluated through both performance and fairness metrics, predominantly involved data collection and preprocessing techniques like resampling and reweighting.

DISCUSSION

This review highlights evolving strategies to mitigate bias in EHR-based AI models, emphasizing the urgent need for both standardized and detailed reporting of the methodologies and systematic real-world testing and evaluation. Such measures are essential for gauging models' practical impact and fostering ethical AI that ensures fairness and equity in healthcare.

Generative models improve fairness of medical classifiers under distribution shifts

Domain generalization is a ubiquitous challenge for machine learning in healthcare. Model performance in real-world conditions might be lower than expected because of discrepancies between the data encountered during deployment and development. Underrepresentation of some groups or conditions during model development is a common cause of this phenomenon. This challenge is often not readily addressed by targeted data acquisition and 'labeling' by expert clinicians, which can be prohibitively expensive or practically impossible because of the rarity of conditions or the available clinical expertise. We hypothesize that advances in generative artificial intelligence can help mitigate this unmet need in a steerable fashion, enriching our training dataset with synthetic examples that address shortfalls of underrepresented conditions or subgroups. We show that diffusion models can automatically learn realistic augmentations from data in a label-efficient manner. We demonstrate that learned augmentations make models more robust and statistically fair in-distribution and out of distribution. To evaluate the generality of our approach, we studied three distinct medical imaging contexts of varying difficulty: (1) histopathology, (2) chest X-ray and (3) dermatology images. Complementing real samples with synthetic ones improved the robustness of models in all three medical tasks and increased fairness by improving the accuracy of clinical diagnosis within underrepresented groups, especially out of distribution.

Drop the shortcuts: image augmentation improves fairness and decreases AI detection of race and other demographics from medical images

BACKGROUND

It has been shown that AI models can learn race on medical images, leading to algorithmic bias. Our aim in this study was to enhance the fairness of medical image models by eliminating bias related to race, age, and sex. We hypothesise models may be learning demographics via shortcut learning and combat this using image augmentation.

METHODS

This study included 44,953 patients who identified as Asian, Black, or White (mean age, 60.68 years ±18.21; 23,499 women) for a total of 194,359 chest X-rays (CXRs) from MIMIC-CXR database. The included CheXpert images comprised 45,095 patients (mean age 63.10 years ±18.14; 20,437 women) for a total of 134,300 CXRs were used for external validation. We also collected 1195 3D brain magnetic resonance imaging (MRI) data from the ADNI database, which included 273 participants with an average age of 76.97 years ±14.22, and 142 females. DL models were trained on either non-augmented or augmented images and assessed using disparity metrics. The features learned by the models were analysed using task transfer experiments and model visualisation techniques.

FINDINGS

In the detection of radiological findings, training a model using augmented CXR images was shown to reduce disparities in error rate among racial groups (-5.45%), age groups (-13.94%), and sex (-22.22%). For AD detection, the model trained with augmented MRI images was shown 53.11% and 31.01% reduction of disparities in error rate among age and sex groups, respectively. Image augmentation led to a reduction in the model's ability to identify demographic attributes and resulted in the model trained for clinical purposes incorporating fewer demographic features.

INTERPRETATION

The model trained using the augmented images was less likely to be influenced by demographic information in detecting image labels. These results demonstrate that the proposed augmentation scheme could enhance the fairness of interpretations by DL models when dealing with data from patients with different demographic backgrounds.

FUNDING

National Science and Technology Council (Taiwan), National Institutes of Health.

Enhancing the fairness of AI prediction models by Quasi-Pareto improvement among heterogeneous thyroid nodule population

Artificial Intelligence (AI) models for medical diagnosis often face challenges of generalizability and fairness. We highlighted the algorithmic unfairness in a large thyroid ultrasound dataset with significant diagnostic performance disparities across subgroups linked causally to sample size imbalances. To address this, we introduced the Quasi-Pareto Improvement (QPI) approach and a deep learning implementation (QP-Net) combining multi-task learning and domain adaptation to improve model performance among disadvantaged subgroups without compromising overall population performance. On the thyroid ultrasound dataset, our method significantly mitigated the area under curve (AUC) disparity for three less-prevalent subgroups by 0.213, 0.112, and 0.173 while maintaining the AUC for dominant subgroups; we also further confirmed the generalizability of our approach on two public datasets: the ISIC2019 skin disease dataset and the CheXpert chest radiograph dataset. Here we show the QPI approach to be widely applicable in promoting AI for equitable healthcare outcomes.

Machine Learning and Bias in Medical Imaging: Opportunities and Challenges

Bias in health care has been well documented and results in disparate and worsened outcomes for at-risk groups. Medical imaging plays a critical role in facilitating patient diagnoses but involves multiple sources of bias including factors related to access to imaging modalities, acquisition of images, and assessment (ie, interpretation) of imaging data. Machine learning (ML) applied to diagnostic imaging has demonstrated the potential to improve the quality of imaging-based diagnosis and the precision of measuring imaging-based traits. Algorithms can leverage subtle information not visible to the human eye to detect underdiagnosed conditions or derive new disease phenotypes by linking imaging features with clinical outcomes, all while mitigating cognitive bias in interpretation. Importantly, however, the application of ML to diagnostic imaging has the potential to either reduce or propagate bias. Understanding the potential gain as well as the potential risks requires an understanding of how and what ML models learn. Common risks of propagating bias can arise from unbalanced training, suboptimal architecture design or selection, and uneven application of models. Notwithstanding these risks, ML may yet be applied to improve gain from imaging across all 3A's (access, acquisition, and assessment) for all patients. In this review, we present a framework for understanding the balance of opportunities and challenges for minimizing bias in medical imaging, how ML may improve current approaches to imaging, and what specific design considerations should be made as part of efforts to maximize the quality of health care for all.

Fairness of artificial intelligence in healthcare: review and recommendations

In this review, we address the issue of fairness in the clinical integration of artificial intelligence (AI) in the medical field. As the clinical adoption of deep learning algorithms, a subfield of AI, progresses, concerns have arisen regarding the impact of AI biases and discrimination on patient health. This review aims to provide a comprehensive overview of concerns associated with AI fairness; discuss strategies to mitigate AI biases; and emphasize the need for cooperation among physicians, AI researchers, AI developers, policymakers, and patients to ensure equitable AI integration. First, we define and introduce the concept of fairness in AI applications in healthcare and radiology, emphasizing the benefits and challenges of incorporating AI into clinical practice. Next, we delve into concerns regarding fairness in healthcare, addressing the various causes of biases in AI and potential concerns such as misdiagnosis, unequal access to treatment, and ethical considerations. We then outline strategies for addressing fairness, such as the importance of diverse and representative data and algorithm audits. Additionally, we discuss ethical and legal considerations such as data privacy, responsibility, accountability, transparency, and explainability in AI. Finally, we present the Fairness of Artificial Intelligence Recommendations in healthcare (FAIR) statement to offer best practices. Through these efforts, we aim to provide a foundation for discussing the responsible and equitable implementation and deployment of AI in healthcare.

Fairness as an afterthought: An American perspective on fairness in model developer-clinician user collaborations

Numerous ethics guidelines have been handed down over the last few years on the ethical applications of machine learning models. Virtually every one of them mentions the importance of "fairness" in the development and use of these models. Unfortunately, though, these ethics documents omit providing a consensually adopted definition or characterization of fairness. As one group of authors observed, these documents treat fairness as an "afterthought" whose importance is undeniable but whose essence seems strikingly elusive. In this essay, which offers a distinctly American treatment of "fairness," we comment on a number of fairness formulations and on qualitative or statistical methods that have been encouraged to achieve fairness. We argue that none of them, at least from an American moral perspective, provides a one-size-fits-all definition of or methodology for securing fairness that could inform or standardize fairness over the universe of use cases witnessing machine learning applications. Instead, we argue that because fairness comprehensions and applications reflect a vast range of use contexts, model developers and clinician users will need to engage in thoughtful collaborations that examine how fairness should be conceived and operationalized in the use case at issue. Part II of this paper illustrates key moments in these collaborations, especially when inter and intra disagreement occurs among model developer and clinician user groups over whether a model is fair or unfair. We conclude by noting that these collaborations will likely occur over the lifetime of a model if its claim to fairness is to advance beyond "afterthought" status.

Longitudinal assessment of demographic representativeness in the Medical Imaging and Data Resource Center open data commons

Purpose

The Medical Imaging and Data Resource Center (MIDRC) open data commons was launched to accelerate the development of artificial intelligence (AI) algorithms to help address the COVID-19 pandemic. The purpose of this study was to quantify longitudinal representativeness of the demographic characteristics of the primary MIDRC dataset compared to the United States general population (US Census) and COVID-19 positive case counts from the Centers for Disease Control and Prevention (CDC).

Approach

The Jensen-Shannon distance (JSD), a measure of similarity of two distributions, was used to longitudinally measure the representativeness of the distribution of (1) all unique patients in the MIDRC data to the 2020 US Census and (2) all unique COVID-19 positive patients in the MIDRC data to the case counts reported by the CDC. The distributions were evaluated in the demographic categories of age at index, sex, race, ethnicity, and the combination of race and ethnicity.

Results

Representativeness of the MIDRC data by ethnicity and the combination of race and ethnicity was impacted by the percentage of CDC case counts for which this was not reported. The distributions by sex and race have retained their level of representativeness over time.

Conclusion

The representativeness of the open medical imaging datasets in the curated public data commons at MIDRC has evolved over time as the number of contributing institutions and overall number of subjects have grown. The use of metrics, such as the JSD support measurement of representativeness, is one step needed for fair and generalizable AI algorithm development.

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.

Generation of a Melanoma and Nevus Data Set From Unstandardized Clinical Photographs on the Internet

Importance

Artificial intelligence (AI) training for diagnosing dermatologic images requires large amounts of clean data. Dermatologic images have different compositions, and many are inaccessible due to privacy concerns, which hinder the development of AI.

Objective

To build a training data set for discriminative and generative AI from unstandardized internet images of melanoma and nevus.

Design, Setting, and Participants

In this diagnostic study, a total of 5619 (CAN5600 data set) and 2006 (CAN2000 data set; a manually revised subset of CAN5600) cropped lesion images of either melanoma or nevus were semiautomatically annotated from approximately 500 000 photographs on the internet using convolutional neural networks (CNNs), region-based CNNs, and large mask inpainting. For unsupervised pretraining, 132 673 possible lesions (LESION130k data set) were also created with diversity by collecting images from 18 482 websites in approximately 80 countries. A total of 5000 synthetic images (GAN5000 data set) were generated using the generative adversarial network (StyleGAN2-ADA; training, CAN2000 data set; pretraining, LESION130k data set).

Main Outcomes and Measures

The area under the receiver operating characteristic curve (AUROC) for determining malignant neoplasms was analyzed. In each test, 1 of the 7 preexisting public data sets (total of 2312 images; including Edinburgh, an SNU subset, Asan test, Waterloo, 7-point criteria evaluation, PAD-UFES-20, and MED-NODE) was used as the test data set. Subsequently, a comparative study was conducted between the performance of the EfficientNet Lite0 CNN on the proposed data set and that trained on the remaining 6 preexisting data sets.

Results

The EfficientNet Lite0 CNN trained on the annotated or synthetic images achieved higher or equivalent mean (SD) AUROCs to the EfficientNet Lite0 trained using the pathologically confirmed public data sets, including CAN5600 (0.874 [0.042]; P = .02), CAN2000 (0.848 [0.027]; P = .08), and GAN5000 (0.838 [0.040]; P = .31 [Wilcoxon signed rank test]) and the preexisting data sets combined (0.809 [0.063]) by the benefits of increased size of the training data set.

Conclusions and Relevance

The synthetic data set in this diagnostic study was created using various AI technologies from internet images. A neural network trained on the created data set (CAN5600) performed better than the same network trained on preexisting data sets combined. Both the annotated (CAN5600 and LESION130k) and synthetic (GAN5000) data sets could be shared for AI training and consensus between physicians.

A dataset of skin lesion images collected in Argentina for the evaluation of AI tools in this population

In recent years, numerous dermatological image databases have been published to make possible the development and validation of artificial intelligence-based technologies to support healthcare professionals in the diagnosis of skin diseases. However, the generation of these datasets confined to certain countries as well as the lack of demographic information accompanying the images, prevents having a real knowledge of in which populations these models could be used. Consequently, this hinders the translation of the models to the clinical setting. This has led the scientific community to encourage the detailed and transparent reporting of the databases used for artificial intelligence developments, as well as to promote the formation of genuinely international databases that can be representative of the world population. Through this work, we seek to provide details of the processing stages of the first public database of dermoscopy and clinical images created in a hospital in Argentina. The dataset comprises 1,616 images corresponding to 1,246 unique lesions collected from 623 patients.

Enhancing Fairness in Disease Prediction by Optimizing Multiple Domain Adversarial Networks

Predictive models in biomedicine need to ensure equitable and reliable outcomes for the populations they are applied to. Unfortunately, biases in medical predictions can lead to unfair treatment and widening disparities, underscoring the need for effective techniques to address these issues. To enhance fairness, we introduce a framework based on a Multiple Domain Adversarial Neural Network (MDANN), which incorporates multiple adversarial components. In an MDANN, an adversarial module is applied to learn a fair pattern by negative gradients back-propagating across multiple sensitive features (i.e., characteristics of individuals that should not be used to discriminate unfairly between individuals when making predictions or decisions.) We leverage loss functions based on the Area Under the Receiver Operating Characteristic Curve (AUC) to address the class imbalance, promoting equitable classification performance for minority groups (e.g., a subset of the population that is underrepresented or disadvantaged.) Moreover, we utilize pre-trained convolutional autoencoders (CAEs) to extract deep representations of data, aiming to enhance prediction accuracy and fairness. Combining these mechanisms, we alleviate biases and disparities to provide reliable and equitable disease prediction. We empirically demonstrate that the MDANN approach leads to better accuracy and fairness in predicting disease progression using brain imaging data for Alzheimer's Disease and Autism populations than state-of-the-art techniques.

Fairness and generalisability in deep learning of retinopathy of prematurity screening algorithms: a literature review

BACKGROUND

Retinopathy of prematurity (ROP) is a vasoproliferative disease responsible for more than 30 000 blind children worldwide. Its diagnosis and treatment are challenging due to the lack of specialists, divergent diagnostic concordance and variation in classification standards. While artificial intelligence (AI) can address the shortage of professionals and provide more cost-effective management, its development needs fairness, generalisability and bias controls prior to deployment to avoid producing harmful unpredictable results. This review aims to compare AI and ROP study's characteristics, fairness and generalisability efforts.

METHODS

Our review yielded 220 articles, of which 18 were included after full-text assessment. The articles were classified into ROP severity grading, plus detection, detecting treatment requiring, ROP prediction and detection of retinal zones.

RESULTS

All the article's authors and included patients are from middle-income and high-income countries, with no low-income countries, South America, Australia and Africa Continents representation.Code is available in two articles and in one on request, while data are not available in any article. 88.9% of the studies use the same retinal camera. In two articles, patients' sex was described, but none applied a bias control in their models.

CONCLUSION

The reviewed articles included 180 228 images and reported good metrics, but fairness, generalisability and bias control remained limited. Reproducibility is also a critical limitation, with few articles sharing codes and none sharing data. Fair and generalisable ROP and AI studies are needed that include diverse datasets, data and code sharing, collaborative research, and bias control to avoid unpredictable and harmful deployments.

Machine learning models for diagnosis and prognosis of Parkinson's disease using brain imaging: general overview, main challenges, and future directions

Parkinson's disease (PD) is a progressive and complex neurodegenerative disorder associated with age that affects motor and cognitive functions. As there is currently no cure, early diagnosis and accurate prognosis are essential to increase the effectiveness of treatment and control its symptoms. Medical imaging, specifically magnetic resonance imaging (MRI), has emerged as a valuable tool for developing support systems to assist in diagnosis and prognosis. The current literature aims to improve understanding of the disease's structural and functional manifestations in the brain. By applying artificial intelligence to neuroimaging, such as deep learning (DL) and other machine learning (ML) techniques, previously unknown relationships and patterns can be revealed in this high-dimensional data. However, several issues must be addressed before these solutions can be safely integrated into clinical practice. This review provides a comprehensive overview of recent ML techniques analyzed for the automatic diagnosis and prognosis of PD in brain MRI. The main challenges in applying ML to medical diagnosis and its implications for PD are also addressed, including current limitations for safe translation into hospitals. These challenges are analyzed at three levels: disease-specific, task-specific, and technology-specific. Finally, potential future directions for each challenge and future perspectives are discussed.

The path toward equal performance in medical machine learning

To ensure equitable quality of care, differences in machine learning model performance between patient groups must be addressed. Here, we argue that two separate mechanisms can cause performance differences between groups. First, model performance may be worse than theoretically achievable in a given group. This can occur due to a combination of group underrepresentation, modeling choices, and the characteristics of the prediction task at hand. We examine scenarios in which underrepresentation leads to underperformance, scenarios in which it does not, and the differences between them. Second, the optimal achievable performance may also differ between groups due to differences in the intrinsic difficulty of the prediction task. We discuss several possible causes of such differences in task difficulty. In addition, challenges such as label biases and selection biases may confound both learning and performance evaluation. We highlight consequences for the path toward equal performance, and we emphasize that leveling up model performance may require gathering not only more data from underperforming groups but also better data. Throughout, we ground our discussion in real-world medical phenomena and case studies while also referencing relevant statistical theory.

Responsible AI practice and AI education are central to AI implementation: a rapid review for all medical imaging professionals in Europe

Artificial intelligence (AI) has transitioned from the lab to the bedside, and it is increasingly being used in healthcare. Radiology and Radiography are on the frontline of AI implementation, because of the use of big data for medical imaging and diagnosis for different patient groups. Safe and effective AI implementation requires that responsible and ethical practices are upheld by all key stakeholders, that there is harmonious collaboration between different professional groups, and customised educational provisions for all involved. This paper outlines key principles of ethical and responsible AI, highlights recent educational initiatives for clinical practitioners and discusses the synergies between all medical imaging professionals as they prepare for the digital future in Europe. Responsible and ethical AI is vital to enhance a culture of safety and trust for healthcare professionals and patients alike. Educational and training provisions for medical imaging professionals on AI is central to the understanding of basic AI principles and applications and there are many offerings currently in Europe. Education can facilitate the transparency of AI tools, but more formalised, university-led training is needed to ensure the academic scrutiny, appropriate pedagogy, multidisciplinarity and customisation to the learners' unique needs are being adhered to. As radiographers and radiologists work together and with other professionals to understand and harness the benefits of AI in medical imaging, it becomes clear that they are faced with the same challenges and that they have the same needs. The digital future belongs to multidisciplinary teams that work seamlessly together, learn together, manage risk collectively and collaborate for the benefit of the patients they serve.

Artificial Intelligence in CT and MR Imaging for Oncological Applications

Cancer care increasingly relies on imaging for patient management. The two most common cross-sectional imaging modalities in oncology are computed tomography (CT) and magnetic resonance imaging (MRI), which provide high-resolution anatomic and physiological imaging. Herewith is a summary of recent applications of rapidly advancing artificial intelligence (AI) in CT and MRI oncological imaging that addresses the benefits and challenges of the resultant opportunities with examples. Major challenges remain, such as how best to integrate AI developments into clinical radiology practice, the vigorous assessment of quantitative CT and MR imaging data accuracy, and reliability for clinical utility and research integrity in oncology. Such challenges necessitate an evaluation of the robustness of imaging biomarkers to be included in AI developments, a culture of data sharing, and the cooperation of knowledgeable academics with vendor scientists and companies operating in radiology and oncology fields. Herein, we will illustrate a few challenges and solutions of these efforts using novel methods for synthesizing different contrast modality images, auto-segmentation, and image reconstruction with examples from lung CT as well as abdome, pelvis, and head and neck MRI. The imaging community must embrace the need for quantitative CT and MRI metrics beyond lesion size measurement. AI methods for the extraction and longitudinal tracking of imaging metrics from registered lesions and understanding the tumor environment will be invaluable for interpreting disease status and treatment efficacy. This is an exciting time to work together to move the imaging field forward with narrow AI-specific tasks. New AI developments using CT and MRI datasets will be used to improve the personalized management of cancer patients.

Human visual explanations mitigate bias in AI-based assessment of surgeon skills

Artificial intelligence (AI) systems can now reliably assess surgeon skills through videos of intraoperative surgical activity. With such systems informing future high-stakes decisions such as whether to credential surgeons and grant them the privilege to operate on patients, it is critical that they treat all surgeons fairly. However, it remains an open question whether surgical AI systems exhibit bias against surgeon sub-cohorts, and, if so, whether such bias can be mitigated. Here, we examine and mitigate the bias exhibited by a family of surgical AI systems-SAIS-deployed on videos of robotic surgeries from three geographically-diverse hospitals (USA and EU). We show that SAIS exhibits an underskilling bias, erroneously downgrading surgical performance, and an overskilling bias, erroneously upgrading surgical performance, at different rates across surgeon sub-cohorts. To mitigate such bias, we leverage a strategy -TWIX-which teaches an AI system to provide a visual explanation for its skill assessment that otherwise would have been provided by human experts. We show that whereas baseline strategies inconsistently mitigate algorithmic bias, TWIX can effectively mitigate the underskilling and overskilling bias while simultaneously improving the performance of these AI systems across hospitals. We discovered that these findings carry over to the training environment where we assess medical students' skills today. Our study is a critical prerequisite to the eventual implementation of AI-augmented global surgeon credentialing programs, ensuring that all surgeons are treated fairly.

Predicting gene mutation status via artificial intelligence technologies based on multimodal integration (MMI) to advance precision oncology

Personalized treatment strategies for cancer frequently rely on the detection of genetic alterations which are determined by molecular biology assays. Historically, these processes typically required single-gene sequencing, next-generation sequencing, or visual inspection of histopathology slides by experienced pathologists in a clinical context. In the past decade, advances in artificial intelligence (AI) technologies have demonstrated remarkable potential in assisting physicians with accurate diagnosis of oncology image-recognition tasks. Meanwhile, AI techniques make it possible to integrate multimodal data such as radiology, histology, and genomics, providing critical guidance for the stratification of patients in the context of precision therapy. Given that the mutation detection is unaffordable and time-consuming for a considerable number of patients, predicting gene mutations based on routine clinical radiological scans or whole-slide images of tissue with AI-based methods has become a hot issue in actual clinical practice. In this review, we synthesized the general framework of multimodal integration (MMI) for molecular intelligent diagnostics beyond standard techniques. Then we summarized the emerging applications of AI in the prediction of mutational and molecular profiles of common cancers (lung, brain, breast, and other tumor types) pertaining to radiology and histology imaging. Furthermore, we concluded that there truly exist multiple challenges of AI techniques in the way of its real-world application in the medical field, including data curation, feature fusion, model interpretability, and practice regulations. Despite these challenges, we still prospect the clinical implementation of AI as a highly potential decision-support tool to aid oncologists in future cancer treatment management.

Artificial intelligence guidance of advanced heart failure therapies: A systematic scoping review

Introduction

Artificial intelligence can recognize complex patterns in large datasets. It is a promising technology to advance heart failure practice, as many decisions rely on expert opinions in the absence of high-quality data-driven evidence.

Methods

We searched Embase, Web of Science, and PubMed databases for articles containing "artificial intelligence," "machine learning," or "deep learning" and any of the phrases "heart transplantation," "ventricular assist device," or "cardiogenic shock" from inception until August 2022. We only included original research addressing post heart transplantation (HTx) or mechanical circulatory support (MCS) clinical care. Review and data extraction were performed in accordance with PRISMA-Scr guidelines.

Results

Of 584 unique publications detected, 31 met the inclusion criteria. The majority focused on outcome prediction post HTx (n = 13) and post durable MCS (n = 7), as well as post HTx and MCS management (n = 7, n = 3, respectively). One study addressed temporary mechanical circulatory support. Most studies advocated for rapid integration of AI into clinical practice, acknowledging potential improvements in management guidance and reliability of outcomes prediction. There was a notable paucity of external data validation and integration of multiple data modalities.

Conclusion

Our review showed mounting innovation in AI application in management of MCS and HTx, with the largest evidence showing improved mortality outcome prediction.

On the benefits of self-taught learning for brain decoding

CONTEXT

We study the benefits of using a large public neuroimaging database composed of functional magnetic resonance imaging (fMRI) statistic maps, in a self-taught learning framework, for improving brain decoding on new tasks. First, we leverage the NeuroVault database to train, on a selection of relevant statistic maps, a convolutional autoencoder to reconstruct these maps. Then, we use this trained encoder to initialize a supervised convolutional neural network to classify tasks or cognitive processes of unseen statistic maps from large collections of the NeuroVault database.

RESULTS

We show that such a self-taught learning process always improves the performance of the classifiers, but the magnitude of the benefits strongly depends on the number of samples available both for pretraining and fine-tuning the models and on the complexity of the targeted downstream task.

CONCLUSION

The pretrained model improves the classification performance and displays more generalizable features, less sensitive to individual differences.

Summary of Year-One Effort of the RCMI Consortium to Enhance Research Capacity and Diversity with Data Science

Despite being disproportionately impacted by health disparities, Black, Hispanic, Indigenous, and other underrepresented populations account for a significant minority of graduates in biomedical data science-related disciplines. Given their commitment to educating underrepresented students and trainees, minority serving institutions (MSIs) can play a significant role in enhancing diversity in the biomedical data science workforce. Little has been published about the reach, curricular breadth, and best practices for delivering these data science training programs. The purpose of this paper is to summarize six Research Centers in Minority Institutions (RCMIs) awarded funding from the National Institute of Minority Health Disparities (NIMHD) to develop new data science training programs. A cross-sectional survey was conducted to better understand the demographics of learners served, curricular topics covered, methods of instruction and assessment, challenges, and recommendations by program directors. Programs demonstrated overall success in reach and curricular diversity, serving a broad range of students and faculty, while also covering a broad range of topics. The main challenges highlighted were a lack of resources and infrastructure and teaching learners with varying levels of experience and knowledge. Further investments in MSIs are needed to sustain training efforts and develop pathways for diversifying the biomedical data science workforce.