A review of explainable and interpretable AI with applications in COVID-19 imaging

Methodology for a fully automated pipeline of AI-based body composition tools for abdominal CT

Accurate, reproducible body composition analysis from abdominal computed tomography (CT) images is critical for both clinical research and patient care. We present a fully automated, artificial intelligence (AI)-based pipeline that streamlines the entire process-from data normalization and anatomical landmarking to automated tissue segmentation and quantitative biomarker extraction. Our methodology ensures standardized inputs and robust segmentation models to compute volumetric, density, and cross-sectional area metrics for a range of organs and tissues. Additionally, we capture selected DICOM header fields to enable downstream analysis of scan parameters and facilitate correction for acquisition-related variability. By emphasizing portability and compatibility across different scanner types, image protocols, and computational environments, we ensure broad applicability of our framework. This toolkit is the basis for the Opportunistic Screening Consortium in Abdominal Radiology (OSCAR) and has been shown to be robust and versatile, critical for large multi-center studies.

The ethics of using artificial intelligence in scientific research: new guidance needed for a new tool

Using artificial intelligence (AI) in research offers many important benefits for science and society but also creates novel and complex ethical issues. While these ethical issues do not necessitate changing established ethical norms of science, they require the scientific community to develop new guidance for the appropriate use of AI. In this article, we briefly introduce AI and explain how it can be used in research, examine some of the ethical issues raised when using it, and offer nine recommendations for responsible use, including: (1) Researchers are responsible for identifying, describing, reducing, and controlling AI-related biases and random errors; (2) Researchers should disclose, describe, and explain their use of AI in research, including its limitations, in language that can be understood by non-experts; (3) Researchers should engage with impacted communities, populations, and other stakeholders concerning the use of AI in research to obtain their advice and assistance and address their interests and concerns, such as issues related to bias; (4) Researchers who use synthetic data should (a) indicate which parts of the data are synthetic; (b) clearly label the synthetic data; (c) describe how the data were generated; and (d) explain how and why the data were used; (5) AI systems should not be named as authors, inventors, or copyright holders but their contributions to research should be disclosed and described; (6) Education and mentoring in responsible conduct of research should include discussion of ethical use of AI.

Discussion of a Simple Method to Generate Descriptive Images Using Predictive ResNet Model Weights and Feature Maps for Recurrent Cervix Cancer

BACKGROUND

Predictive models like Residual Neural Networks (ResNets) can use Magnetic Resonance Imaging (MRI) data to identify cervix tumors likely to recur after radiotherapy (RT) with high accuracy. However, there persists a lack of insight into model selections (explainability). In this study, we explored whether model features could be used to generate simulated images as a method of model explainability.

METHODS

T2W MRI data were collected for twenty-seven women with cervix cancer who received RT from the TCGA-CESC database. Simulated images were generated as follows: [A] a ResNet model was trained to identify recurrent cervix cancer; [B] a model was evaluated on T2W MRI data for subjects to obtain corresponding feature maps; [C] most important feature maps were determined for each image; [D] feature maps were combined across all images to generate a simulated image; [E] the final image was reviewed by a radiation oncologist and an initial algorithm to identify the likelihood of recurrence.

RESULTS

Predictive feature maps from the ResNet model (93% accuracy) were used to generate simulated images. Simulated images passed through the model were identified as recurrent and non-recurrent cervix tumors after radiotherapy. A radiation oncologist identified the simulated images as cervix tumors with characteristics of aggressive Cervical Cancer. These images also contained multiple MRI features not considered clinically relevant.

CONCLUSION

This simple method was able to generate simulated MRI data that mimicked recurrent and non-recurrent cervix cancer tumor images. These generated images could be useful for evaluating the explainability of predictive models and to assist radiologists with the identification of features likely to predict disease course.

Integrating artificial intelligence with smartphone-based imaging for cancer detection in vivo

Cancer is a major global health challenge, accounting for nearly one in six deaths worldwide. Early diagnosis significantly improves survival rates and patient outcomes, yet in resource-limited settings, the scarcity of medical resources often leads to late-stage diagnosis. Integrating artificial intelligence (AI) with smartphone-based imaging systems offers a promising solution by providing portable, cost-effective, and widely accessible tools for early cancer detection. This paper introduces advanced smartphone-based imaging systems that utilize various imaging modalities for in vivo detection of different cancer types and highlights the advancements of AI for in vivo cancer detection in smartphone-based imaging. However, these compact smartphone systems face challenges like low imaging quality and restricted computing power. The use of advanced AI algorithms to address the optical and computational limitations of smartphone-based imaging systems provides promising solutions. AI-based cancer detection also faces challenges. Transparency and reliability are critical factors in gaining the trust and acceptance of AI algorithms for clinical application, explainable and uncertainty-aware AI breaks the black box and will shape the future AI development in early cancer detection. The challenges and solutions for improving AI accuracy, transparency, and reliability are general issues in AI applications, the AI technologies, limitations, and potentials discussed in this paper are applicable to a wide range of biomedical imaging diagnostics beyond smartphones or cancer-specific applications. Smartphone-based multimodal imaging systems and deep learning algorithms for multimodal data analysis are also growing trends, as this approach can provide comprehensive information about the tissue being examined. Future opportunities and perspectives of AI-integrated smartphone imaging systems will be to make cutting-edge diagnostic tools more affordable and accessible, ultimately enabling early cancer detection for a broader population.

Local interpretable model-agnostic explanation approach for medical imaging analysis: A systematic literature review

BACKGROUND

The interpretability and explainability of machine learning (ML) and artificial intelligence systems are critical for generating trust in their outcomes in fields such as medicine and healthcare. Errors generated by these systems, such as inaccurate diagnoses or treatments, can have serious and even life-threatening effects on patients. Explainable Artificial Intelligence (XAI) is emerging as an increasingly significant area of research nowadays, focusing on the black-box aspect of sophisticated and difficult-to-interpret ML algorithms. XAI techniques such as Local Interpretable Model-Agnostic Explanations (LIME) can give explanations for these models, raising confidence in the systems and improving trust in their predictions. Numerous works have been published that respond to medical problems through the use of ML models in conjunction with XAI algorithms to give interpretability and explainability. The primary objective of the study is to evaluate the performance of the newly emerging LIME techniques within healthcare domains that require more attention in the realm of XAI research.

METHOD

A systematic search was conducted in numerous databases (Scopus, Web of Science, IEEE Xplore, ScienceDirect, MDPI, and PubMed) that identified 1614 peer-reviewed articles published between 2019 and 2023.

RESULTS

52 articles were selected for detailed analysis that showed a growing trend in the application of LIME techniques in healthcare, with significant improvements in the interpretability of ML models used for diagnostic and prognostic purposes.

CONCLUSION

The findings suggest that the integration of XAI techniques, particularly LIME, enhances the transparency and trustworthiness of AI systems in healthcare, thereby potentially improving patient outcomes and fostering greater acceptance of AI-driven solutions among medical professionals.

Anatomical prior-based vertebral landmark detection for spinal disorder diagnosis

As one of fundamental ways to interpret spine images, detection of vertebral landmarks is an informative prerequisite for further diagnosis and management of spine disorders such as scoliosis and fractures. Most existing machine learning-based methods for automatic vertebral landmark detection suffer from overlapping landmarks or abnormally long distances between nearby landmarks against anatomical priors, and thus lack sufficient reliability and interpretability. To tackle the problem, this paper systematically utilizes anatomical prior knowledge in vertebral landmark detection. We explicitly formulate anatomical priors of the spine, related to distances among vertebrae and spatial order within the spine, and integrate these geometrical constraints within training loss, inference procedure, and evaluation metrics. First, we introduce an anatomy-constraint loss to regularize the training process with the aforementioned contextual priors explicitly. Second, we propose a simple-yet-effective anatomy-aided inference procedure by employing sequential prediction rather than a parallel counterpart. Third, we provide novel anatomy-related metrics to quantitatively evaluate to which extent landmark predictions follow the anatomical priors, as is not reflected within the widely-used landmark localization error metric. We employ the localization framework on 1410 anterior-posterior radiographic images. Compared with competitive baseline models, we achieve superior landmark localization accuracy and comparable Cobb angle estimation for scoliosis assessment. Ablation studies demonstrate the effectiveness of designed components on the decrease of localization error and improvement of anatomical plausibility. Additionally, we exhibit effective generalization performance by transferring our detection method onto sagittal 2-D slices of CT scans and boost the performance of downstream compression fracture classification at vertebra-level.

Enhancing interpretability and accuracy of AI models in healthcare: a comprehensive review on challenges and future directions

Artificial Intelligence (AI) has demonstrated exceptional performance in automating critical healthcare tasks, such as diagnostic imaging analysis and predictive modeling, often surpassing human capabilities. The integration of AI in healthcare promises substantial improvements in patient outcomes, including faster diagnosis and personalized treatment plans. However, AI models frequently lack interpretability, leading to significant challenges concerning their performance and generalizability across diverse patient populations. These opaque AI technologies raise serious patient safety concerns, as non-interpretable models can result in improper treatment decisions due to misinterpretations by healthcare providers. Our systematic review explores various AI applications in healthcare, focusing on the critical assessment of model interpretability and accuracy. We identify and elucidate the most significant limitations of current AI systems, such as the black-box nature of deep learning models and the variability in performance across different clinical settings. By addressing these challenges, our objective is to provide healthcare providers with well-informed strategies to develop innovative and safe AI solutions. This review aims to ensure that future AI implementations in healthcare not only enhance performance but also maintain transparency and patient safety.

Simulating clinical features on chest radiographs for medical image exploration and CNN explainability using a style-based generative adversarial autoencoder

Explainability of convolutional neural networks (CNNs) is integral for their adoption into radiological practice. Commonly used attribution methods localize image areas important for CNN prediction but do not characterize relevant imaging features underlying these areas, acting as a barrier to the adoption of CNNs for clinical use. We therefore propose Semantic Exploration and Explainability using a Style-based Generative Adversarial Autoencoder Network (SEE-GAAN), an explainability framework that uses latent space manipulation to generate a sequence of synthetic images that semantically visualizes how clinical and CNN features manifest within medical images. Visual analysis of changes in these sequences then facilitates the interpretation of features, thereby improving explainability. SEE-GAAN was first developed on a cohort of 26,664 chest radiographs across 15,409 patients from our institution. SEE-GAAN sequences were then generated across several clinical features and CNN predictions of NT-pro B-type natriuretic peptide (BNPP) as a proxy for acute heart failure. Radiological interpretations indicated SEE-GAAN sequences captured relevant changes in anatomical and pathological morphology associated with clinical and CNN predictions and clarified ambiguous areas highlighted by commonly used attribution methods. Our study demonstrates SEE-GAAN can facilitate our understanding of clinical features for imaging biomarker exploration and improve CNN transparency over commonly used explainability methods.

A joint ESTRO and AAPM guideline for development, clinical validation and reporting of artificial intelligence models in radiation therapy

BACKGROUND AND PURPOSE

Artificial Intelligence (AI) models in radiation therapy are being developed with increasing pace. Despite this, the radiation therapy community has not widely adopted these models in clinical practice. A cohesive guideline on how to develop, report and clinically validate AI algorithms might help bridge this gap.

METHODS AND MATERIALS

A Delphi process with all co-authors was followed to determine which topics should be addressed in this comprehensive guideline. Separate sections of the guideline, including Statements, were written by subgroups of the authors and discussed with the whole group at several meetings. Statements were formulated and scored as highly recommended or recommended.

RESULTS

The following topics were found most relevant: Decision making, image analysis, volume segmentation, treatment planning, patient specific quality assurance of treatment delivery, adaptive treatment, outcome prediction, training, validation and testing of AI model parameters, model availability for others to verify, model quality assurance/updates and upgrades, ethics. Key references were given together with an outlook on current hurdles and possibilities to overcome these. 19 Statements were formulated.

CONCLUSION

A cohesive guideline has been written which addresses main topics regarding AI in radiation therapy. It will help to guide development, as well as transparent and consistent reporting and validation of new AI tools and facilitate adoption.

The AI Future of Emergency Medicine

In the coming years, artificial intelligence (AI) and machine learning will likely give rise to profound changes in the field of emergency medicine, and medicine more broadly. This article discusses these anticipated changes in terms of 3 overlapping yet distinct stages of AI development. It reviews some fundamental concepts in AI and explores their relation to clinical practice, with a focus on emergency medicine. In addition, it describes some of the applications of AI in disease diagnosis, prognosis, and treatment, as well as some of the practical issues that they raise, the barriers to their implementation, and some of the legal and regulatory challenges they create.

Deciphering the Feature Representation of Deep Neural Networks for High-Performance AI

AI driven by deep learning is transforming many aspects of science and technology. The enormous success of deep learning stems from its unique capability of extracting essential features from Big Data for decision-making. However, the feature extraction and hidden representations in deep neural networks (DNNs) remain inexplicable, primarily because of lack of technical tools to comprehend and interrogate the feature space data. The main hurdle here is that the feature data are often noisy in nature, complex in structure, and huge in size and dimensionality, making it intractable for existing techniques to analyze the data reliably. In this work, we develop a computational framework named contrastive feature analysis (CFA) to facilitate the exploration of the DNN feature space and improve the performance of AI. By utilizing the interaction relations among the features and incorporating a novel data-driven kernel formation strategy into the feature analysis pipeline, CFA mitigates the limitations of traditional approaches and provides an urgently needed solution for the analysis of feature space data. The technique allows feature data exploration in unsupervised, semi-supervised and supervised formats to address different needs of downstream applications. The potential of CFA and its applications for pruning of neural network architectures are demonstrated using several state-of-the-art networks and well-annotated datasets across different disciplines.

Deep Learning-Guided Dosimetry for Mitigating Local Failure of Patients With Non-Small Cell Lung Cancer Receiving Stereotactic Body Radiation Therapy

PURPOSE

Non-small cell lung cancer (NSCLC) stereotactic body radiation therapy with 50 Gy/5 fractions is sometimes considered controversial, as the nominal biologically effective dose (BED) of 100 Gy is felt by some to be insufficient for long-term local control of some lesions. In this study, we analyzed such patients using explainable deep learning techniques and consequently proposed appropriate treatment planning criteria. These novel criteria could help planners achieve optimized treatment plans for maximal local control.

METHODS AND MATERIALS

A total of 535 patients treated with 50 Gy/5 fractions were used to develop a novel deep learning local response model. A multimodality approach, incorporating computed tomography images, 3-dimensional dose distribution, and patient demographics, combined with a discrete-time survival model, was applied to predict time to failure and the probability of local control. Subsequently, an integrated gradient-weighted class activation mapping method was used to identify the most significant dose-volume metrics predictive of local failure and their optimal cut-points.

RESULTS

The model was cross-validated, showing an acceptable performance (c-index: 0.72, 95% CI, 0.68-0.75); the testing c-index was 0.69. The model's spatial attention was concentrated mostly in the tumors' periphery (planning target volume [PTV] - internal gross target volume [IGTV]) region. Statistically significant dose-volume metrics in improved local control were BED Dnear-min ≥ 103.8 Gy in IGTV (hazard ratio [HR], 0.31; 95% CI, 015-0.63), V104 ≥ 98% in IGTV (HR, 0.30; 95% CI, 0.15-0.60), gEUD ≥ 103.8 Gy in PTV-IGTV (HR, 0.25; 95% CI, 0.12-0.50), and Dmean ≥ 104.5 Gy in PTV-IGTV (HR, 0.25; 95% CI, 0.12-0.51).

CONCLUSIONS

Deep learning-identified dose-volume metrics have shown significant prognostic power (log-rank, P = .003) and could be used as additional actionable criteria for treatment planning in NSCLC stereotactic body radiation therapy patients receiving 50 Gy in 5 fractions. Although our data do not confirm or refute that a significantly higher BED for the prescription dose is necessary for tumor control in NSCLC, it might be clinically effective to escalate the nominal prescribed dose from BED 100 to 105 Gy.

Robust explanation supervision for false positive reduction in pulmonary nodule detection

BACKGROUND

Lung cancer is the deadliest and second most common cancer in the United States due to the lack of symptoms for early diagnosis. Pulmonary nodules are small abnormal regions that can be potentially correlated to the occurrence of lung cancer. Early detection of these nodules is critical because it can significantly improve the patient's survival rates. Thoracic thin-sliced computed tomography (CT) scanning has emerged as a widely used method for diagnosing and prognosis lung abnormalities.

PURPOSE

The standard clinical workflow of detecting pulmonary nodules relies on radiologists to analyze CT images to assess the risk factors of cancerous nodules. However, this approach can be error-prone due to the various nodule formation causes, such as pollutants and infections. Deep learning (DL) algorithms have recently demonstrated remarkable success in medical image classification and segmentation. As an ever more important assistant to radiologists in nodule detection, it is imperative ensure the DL algorithm and radiologist to better understand the decisions from each other. This study aims to develop a framework integrating explainable AI methods to achieve accurate pulmonary nodule detection.

METHODS

A robust and explainable detection (RXD) framework is proposed, focusing on reducing false positives in pulmonary nodule detection. Its implementation is based on an explanation supervision method, which uses nodule contours of radiologists as supervision signals to force the model to learn nodule morphologies, enabling improved learning ability on small dataset, and enable small dataset learning ability. In addition, two imputation methods are applied to the nodule region annotations to reduce the noise within human annotations and allow the model to have robust attributions that meet human expectations. The 480, 265, and 265 CT image sets from the public Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) dataset are used for training, validation, and testing.

RESULTS

Using only 10, 30, 50, and 100 training samples sequentially, our method constantly improves the classification performance and explanation quality of baseline in terms of Area Under the Curve (AUC) and Intersection over Union (IoU). In particular, our framework with a learnable imputation kernel improves IoU from baseline by 24.0% to 80.0%. A pre-defined Gaussian imputation kernel achieves an even greater improvement, from 38.4% to 118.8% from baseline. Compared to the baseline trained on 100 samples, our method shows less drop in AUC when trained on fewer samples. A comprehensive comparison of interpretability shows that our method aligns better with expert opinions.

CONCLUSIONS

A pulmonary nodule detection framework was demonstrated using public thoracic CT image datasets. The framework integrates the robust explanation supervision (RES) technique to ensure the performance of nodule classification and morphology. The method can reduce the workload of radiologists and enable them to focus on the diagnosis and prognosis of the potential cancerous pulmonary nodules at the early stage to improve the outcomes for lung cancer patients.

A critical moment in machine learning in medicine: on reproducible and interpretable learning

Over the past two decades, advances in computational power and data availability combined with increased accessibility to pre-trained models have led to an exponential rise in machine learning (ML) publications. While ML may have the potential to transform healthcare, this sharp increase in ML research output without focus on methodological rigor and standard reporting guidelines has fueled a reproducibility crisis. In addition, the rapidly growing complexity of these models compromises their interpretability, which currently impedes their successful and widespread clinical adoption. In medicine, where failure of such models may have severe implications for patients' health, the high requirements for accuracy, robustness, and interpretability confront ML researchers with a unique set of challenges. In this review, we discuss the semantics of reproducibility and interpretability, as well as related issues and challenges, and outline possible solutions to counteracting the "black box". To foster reproducibility, standard reporting guidelines need to be further developed and data or code sharing encouraged. Editors and reviewers may equally play a critical role by establishing high methodological standards and thus preventing the dissemination of low-quality ML publications. To foster interpretable learning, the use of simpler models more suitable for medical data can inform the clinician how results are generated based on input data. Model-agnostic explanation tools, sensitivity analysis, and hidden layer representations constitute further promising approaches to increase interpretability. Balancing model performance and interpretability are important to ensure clinical applicability. We have now reached a critical moment for ML in medicine, where addressing these issues and implementing appropriate solutions will be vital for the future evolution of the field.

Transformative Potential of AI in Healthcare: Definitions, Applications, and Navigating the Ethical Landscape and Public Perspectives

Artificial intelligence (AI) has emerged as a crucial tool in healthcare with the primary aim of improving patient outcomes and optimizing healthcare delivery. By harnessing machine learning algorithms, natural language processing, and computer vision, AI enables the analysis of complex medical data. The integration of AI into healthcare systems aims to support clinicians, personalize patient care, and enhance population health, all while addressing the challenges posed by rising costs and limited resources. As a subdivision of computer science, AI focuses on the development of advanced algorithms capable of performing complex tasks that were once reliant on human intelligence. The ultimate goal is to achieve human-level performance with improved efficiency and accuracy in problem-solving and task execution, thereby reducing the need for human intervention. Various industries, including engineering, media/entertainment, finance, and education, have already reaped significant benefits by incorporating AI systems into their operations. Notably, the healthcare sector has witnessed rapid growth in the utilization of AI technology. Nevertheless, there remains untapped potential for AI to truly revolutionize the industry. It is important to note that despite concerns about job displacement, AI in healthcare should not be viewed as a threat to human workers. Instead, AI systems are designed to augment and support healthcare professionals, freeing up their time to focus on more complex and critical tasks. By automating routine and repetitive tasks, AI can alleviate the burden on healthcare professionals, allowing them to dedicate more attention to patient care and meaningful interactions. However, legal and ethical challenges must be addressed when embracing AI technology in medicine, alongside comprehensive public education to ensure widespread acceptance.

Trust and stakeholder perspectives on the implementation of AI tools in clinical radiology

OBJECTIVES

To define requirements that condition trust in artificial intelligence (AI) as clinical decision support in radiology from the perspective of various stakeholders and to explore ways to fulfil these requirements.

METHODS

Semi-structured interviews were conducted with twenty-five respondents-nineteen directly involved in the development, implementation, or use of AI applications in radiology and six working with AI in other areas of healthcare. We designed the questions to explore three themes: development and use of AI, professional decision-making, and management and organizational procedures connected to AI. The transcribed interviews were analysed in an iterative coding process from open coding to theoretically informed thematic coding.

RESULTS

We identified four aspects of trust that relate to reliability, transparency, quality verification, and inter-organizational compatibility. These aspects fall under the categories of substantial and procedural requirements.

CONCLUSIONS

Development of appropriate levels of trust in AI in healthcare is complex and encompasses multiple dimensions of requirements. Various stakeholders will have to be involved in developing AI solutions for healthcare and radiology to fulfil these requirements.

CLINICAL RELEVANCE STATEMENT

For AI to achieve advances in radiology, it must be given the opportunity to support, rather than replace, human expertise. Support requires trust. Identification of aspects and conditions for trust allows developing AI implementation strategies that facilitate advancing the field.

KEY POINTS

• Dimensions of procedural and substantial demands that need to be fulfilled to foster appropriate levels of trust in AI in healthcare are conditioned on aspects related to reliability, transparency, quality verification, and inter-organizational compatibility. •Creating the conditions for trust to emerge requires the involvement of various stakeholders, who will have to compensate the problem's inherent complexity by finding and promoting well-defined solutions.

Design of an interface to communicate artificial intelligence-based prognosis for patients with advanced solid tumors: a user-centered approach

OBJECTIVES

To design an interface to support communication of machine learning (ML)-based prognosis for patients with advanced solid tumors, incorporating oncologists' needs and feedback throughout design.

MATERIALS AND METHODS

Using an interdisciplinary user-centered design approach, we performed 5 rounds of iterative design to refine an interface, involving expert review based on usability heuristics, input from a color-blind adult, and 13 individual semi-structured interviews with oncologists. Individual interviews included patient vignettes and a series of interfaces populated with representative patient data and predicted survival for each treatment decision point when a new line of therapy (LoT) was being considered. Ongoing feedback informed design decisions, and directed qualitative content analysis of interview transcripts was used to evaluate usability and identify enhancement requirements.

RESULTS

Design processes resulted in an interface with 7 sections, each addressing user-focused questions, supporting oncologists to "tell a story" as they discuss prognosis during a clinical encounter. The iteratively enhanced interface both triggered and reflected design decisions relevant when attempting to communicate ML-based prognosis, and exposed misassumptions. Clinicians requested enhancements that emphasized interpretability over explainability. Qualitative findings confirmed that previously identified issues were resolved and clarified necessary enhancements (eg, use months not days) and concerns about usability and trust (eg, address LoT received elsewhere). Appropriate use should be in the context of a conversation with an oncologist.

CONCLUSION

User-centered design, ongoing clinical input, and a visualization to communicate ML-related outcomes are important elements for designing any decision support tool enabled by artificial intelligence, particularly when communicating prognosis risk.

Prediction of oxygen supplementation by a deep-learning model integrating clinical parameters and chest CT images in COVID-19

PURPOSE

As of March 2023, the number of patients with COVID-19 worldwide is declining, but the early diagnosis of patients requiring inpatient treatment and the appropriate allocation of limited healthcare resources remain unresolved issues. In this study we constructed a deep-learning (DL) model to predict the need for oxygen supplementation using clinical information and chest CT images of patients with COVID-19.

MATERIALS AND METHODS

We retrospectively enrolled 738 patients with COVID-19 for whom clinical information (patient background, clinical symptoms, and blood test findings) was available and chest CT imaging was performed. The initial data set was divided into 591 training and 147 evaluation data. We developed a DL model that predicted oxygen supplementation by integrating clinical information and CT images. The model was validated at two other facilities (n = 191 and n = 230). In addition, the importance of clinical information for prediction was assessed.

RESULTS

The proposed DL model showed an area under the curve (AUC) of 89.9% for predicting oxygen supplementation. Validation from the two other facilities showed an AUC > 80%. With respect to interpretation of the model, the contribution of dyspnea and the lactate dehydrogenase level was higher in the model.

CONCLUSIONS

The DL model integrating clinical information and chest CT images had high predictive accuracy. DL-based prediction of disease severity might be helpful in the clinical management of patients with COVID-19.

Sequestration of imaging studies in MIDRC: stratified sampling to balance demographic characteristics of patients in a multi-institutional data commons

Purpose

The Medical Imaging and Data Resource Center (MIDRC) is a multi-institutional effort to accelerate medical imaging machine intelligence research and create a publicly available image repository/commons as well as a sequestered commons for performance evaluation and benchmarking of algorithms. After de-identification, approximately 80% of the medical images and associated metadata become part of the open commons and 20% are sequestered from the open commons. To ensure that both commons are representative of the population available, we introduced a stratified sampling method to balance the demographic characteristics across the two datasets.

Approach

Our method uses multi-dimensional stratified sampling where several demographic variables of interest are sequentially used to separate the data into individual strata, each representing a unique combination of variables. Within each resulting stratum, patients are assigned to the open or sequestered commons. This algorithm was used on an example dataset containing 5000 patients using the variables of race, age, sex at birth, ethnicity, COVID-19 status, and image modality and compared resulting demographic distributions to naïve random sampling of the dataset over 2000 independent trials.

Results

Resulting prevalence of each demographic variable matched the prevalence from the input dataset within one standard deviation. Mann-Whitney U test results supported the hypothesis that sequestration by stratified sampling provided more balanced subsets than naïve randomization, except for demographic subcategories with very low prevalence.

Conclusions

The developed multi-dimensional stratified sampling algorithm can partition a large dataset while maintaining balance across several variables, superior to the balance achieved from naïve randomization.

Translation of AI into oncology clinical practice

Artificial intelligence (AI) is a transformative technology that is capturing popular imagination and can revolutionize biomedicine. AI and machine learning (ML) algorithms have the potential to break through existing barriers in oncology research and practice such as automating workflow processes, personalizing care, and reducing healthcare disparities. Emerging applications of AI/ML in the literature include screening and early detection of cancer, disease diagnosis, response prediction, prognosis, and accelerated drug discovery. Despite this excitement, only few AI/ML models have been properly validated and fewer have become regulated products for routine clinical use. In this review, we highlight the main challenges impeding AI/ML clinical translation. We present different clinical use cases from the domains of radiology, radiation oncology, immunotherapy, and drug discovery in oncology. We dissect the unique challenges and opportunities associated with each of these cases. Finally, we summarize the general requirements for successful AI/ML implementation in the clinic, highlighting specific examples and points of emphasis including the importance of multidisciplinary collaboration of stakeholders, role of domain experts in AI augmentation, transparency of AI/ML models, and the establishment of a comprehensive quality assurance program to mitigate risks of training bias and data drifts, all culminating toward safer and more beneficial AI/ML applications in oncology labs and clinics.

AI and machine learning ethics, law, diversity, and global impact

Artificial intelligence (AI) and its machine learning (ML) algorithms are offering new promise for personalized biomedicine and more cost-effective healthcare with impressive technical capability to mimic human cognitive capabilities. However, widespread application of this promising technology has been limited in the medical domain and expectations have been tampered by ethical challenges and concerns regarding patient privacy, legal responsibility, trustworthiness, and fairness. To balance technical innovation with ethical applications of AI/ML, developers must demonstrate the AI functions as intended and adopt strategies to minimize the risks for failure or bias. This review describes the new ethical challenges created by AI/ML for clinical care and identifies specific considerations for its practice in medicine. We provide an overview of regulatory and legal issues applicable in Europe and the United States, a description of technical aspects to consider, and present recommendations for trustworthy AI/ML that promote transparency, minimize risks of bias or error, and protect the patient well-being.

Covid-19 triage in the emergency department 2.0: how analytics and AI transform a human-made algorithm for the prediction of clinical pathways

The Covid-19 pandemic has pushed many hospitals to their capacity limits. Therefore, a triage of patients has been discussed controversially primarily through an ethical perspective. The term triage contains many aspects such as urgency of treatment, severity of the disease and pre-existing conditions, access to critical care, or the classification of patients regarding subsequent clinical pathways starting from the emergency department. The determination of the pathways is important not only for patient care, but also for capacity planning in hospitals. We examine the performance of a human-made triage algorithm for clinical pathways which is considered a guideline for emergency departments in Germany based on a large multicenter dataset with over 4,000 European Covid-19 patients from the LEOSS registry. We find an accuracy of 28 percent and approximately 15 percent sensitivity for the ward class. The results serve as a benchmark for our extensions including an additional category of palliative care as a new label, analytics, AI, XAI, and interactive techniques. We find significant potential of analytics and AI in Covid-19 triage regarding accuracy, sensitivity, and other performance metrics whilst our interactive human-AI algorithm shows superior performance with approximately 73 percent accuracy and up to 76 percent sensitivity. The results are independent of the data preparation process regarding the imputation of missing values or grouping of comorbidities. In addition, we find that the consideration of an additional label palliative care does not improve the results.

How does the model make predictions? A systematic literature review on the explainability power of machine learning in healthcare

BACKGROUND

Medical use cases for machine learning (ML) are growing exponentially. The first hospitals are already using ML systems as decision support systems in their daily routine. At the same time, most ML systems are still opaque and it is not clear how these systems arrive at their predictions.

METHODS

In this paper, we provide a brief overview of the taxonomy of explainability methods and review popular methods. In addition, we conduct a systematic literature search on PubMed to investigate which explainable artificial intelligence (XAI) methods are used in 450 specific medical supervised ML use cases, how the use of XAI methods has emerged recently, and how the precision of describing ML pipelines has evolved over the past 20 years.

RESULTS

A large fraction of publications with ML use cases do not use XAI methods at all to explain ML predictions. However, when XAI methods are used, open-source and model-agnostic explanation methods are more commonly used, with SHapley Additive exPlanations (SHAP) and Gradient Class Activation Mapping (Grad-CAM) for tabular and image data leading the way. ML pipelines have been described in increasing detail and uniformity in recent years. However, the willingness to share data and code has stagnated at about one-quarter.

CONCLUSIONS

XAI methods are mainly used when their application requires little effort. The homogenization of reports in ML use cases facilitates the comparability of work and should be advanced in the coming years. Experts who can mediate between the worlds of informatics and medicine will become more and more in demand when using ML systems due to the high complexity of the domain.

Genetic data visualization using literature text-based neural networks: Examples associated with myocardial infarction

Data visualization is critical to unraveling hidden information from complex and high-dimensional data. Interpretable visualization methods are critical, especially in the biology and medical fields, however, there are limited effective visualization methods for large genetic data. Current visualization methods are limited to lower-dimensional data and their performance suffers if there is missing data. In this study, we propose a literature-based visualization method to reduce high-dimensional data without compromising the dynamics of the single nucleotide polymorphisms (SNP) and textual interpretability. Our method is innovative because it is shown to (1) preserves both global and local structures of SNP while reducing the dimension of the data using literature text representations, and (2) enables interpretable visualizations using textual information. For performance evaluations, we examined the proposed approach to classify various classification categories including race, myocardial infarction event age groups, and sex using several machine learning models on the literature-derived SNP data. We used visualization approaches to examine clustering of data as well as quantitative performance metrics for the classification of the risk factors examined above. Our method outperformed all popular dimensionality reduction and visualization methods for both classification and visualization, and it is robust against missing and higher-dimensional data. Moreover, we found it feasible to incorporate both genetic and other risk information obtained from literature with our method.

Explainable AI in radiology: a white paper of the Italian Society of Medical and Interventional Radiology

The term Explainable Artificial Intelligence (xAI) groups together the scientific body of knowledge developed while searching for methods to explain the inner logic behind the AI algorithm and the model inference based on knowledge-based interpretability. The xAI is now generally recognized as a core area of AI. A variety of xAI methods currently are available to researchers; nonetheless, the comprehensive classification of the xAI methods is still lacking. In addition, there is no consensus among the researchers with regards to what an explanation exactly is and which are salient properties that must be considered to make it understandable for every end-user. The SIRM introduces an xAI-white paper, which is intended to aid Radiologists, medical practitioners, and scientists in the understanding an emerging field of xAI, the black-box problem behind the success of the AI, the xAI methods to unveil the black-box into a glass-box, the role, and responsibilities of the Radiologists for appropriate use of the AI-technology. Due to the rapidly changing and evolution of AI, a definitive conclusion or solution is far away from being defined. However, one of our greatest responsibilities is to keep up with the change in a critical manner. In fact, ignoring and discrediting the advent of AI a priori will not curb its use but could result in its application without awareness. Therefore, learning and increasing our knowledge about this very important technological change will allow us to put AI at our service and at the service of the patients in a conscious way, pushing this paradigm shift as far as it will benefit us.

Explainable artificial intelligence (XAI) in radiology and nuclear medicine: a literature review

Rational

Deep learning (DL) has demonstrated a remarkable performance in diagnostic imaging for various diseases and modalities and therefore has a high potential to be used as a clinical tool. However, current practice shows low deployment of these algorithms in clinical practice, because DL algorithms lack transparency and trust due to their underlying black-box mechanism. For successful employment, explainable artificial intelligence (XAI) could be introduced to close the gap between the medical professionals and the DL algorithms. In this literature review, XAI methods available for magnetic resonance (MR), computed tomography (CT), and positron emission tomography (PET) imaging are discussed and future suggestions are made.

Methods

PubMed, Embase.com and Clarivate Analytics/Web of Science Core Collection were screened. Articles were considered eligible for inclusion if XAI was used (and well described) to describe the behavior of a DL model used in MR, CT and PET imaging.

Results

A total of 75 articles were included of which 54 and 17 articles described post and ad hoc XAI methods, respectively, and 4 articles described both XAI methods. Major variations in performance is seen between the methods. Overall, post hoc XAI lacks the ability to provide class-discriminative and target-specific explanation. Ad hoc XAI seems to tackle this because of its intrinsic ability to explain. However, quality control of the XAI methods is rarely applied and therefore systematic comparison between the methods is difficult.

Conclusion

There is currently no clear consensus on how XAI should be deployed in order to close the gap between medical professionals and DL algorithms for clinical implementation. We advocate for systematic technical and clinical quality assessment of XAI methods. Also, to ensure end-to-end unbiased and safe integration of XAI in clinical workflow, (anatomical) data minimization and quality control methods should be included.

MIDRC CRP10 AI interface-an integrated tool for exploring, testing and visualization of AI models

Objective. Developing Machine Learning models (N Gorre et al 2023) for clinical applications from scratch can be a cumbersome task requiring varying levels of expertise. Seasoned developers and researchers may also often face incompatible frameworks and data preparation issues. This is further complicated in the context of diagnostic radiology and oncology applications, given the heterogenous nature of the input data and the specialized task requirements. Our goal is to provide clinicians, researchers, and early AI developers with a modular, flexible, and user-friendly software tool that can effectively meet their needs to explore, train, and test AI algorithms by allowing users to interpret their model results. This latter step involves the incorporation of interpretability and explainability methods that would allow visualizing performance as well as interpreting predictions across the different neural network layers of a deep learning algorithm.Approach. To demonstrate our proposed tool, we have developed the CRP10 AI Application Interface (CRP10AII) as part of the MIDRC consortium. CRP10AII is based on the web service Django framework in Python. CRP10AII/Django/Python in combination with another data manager tool/platform, data commons such as Gen3 can provide a comprehensive while easy to use machine/deep learning analytics tool. The tool allows to test, visualize, interpret how and why the deep learning model is performing. The major highlight of CRP10AII is its capability of visualization and interpretability of otherwise Blackbox AI algorithms.Results. CRP10AII provides many convenient features for model building and evaluation, including: (1) query and acquire data according to the specific application (e.g. classification, segmentation) from the data common platform (Gen3 here); (2) train the AI models from scratch or use pre-trained models (e.g. VGGNet, AlexNet, BERT) for transfer learning and test the model predictions, performance assessment, receiver operating characteristics curve evaluation; (3) interpret the AI model predictions using methods like SHAPLEY, LIME values; and (4) visualize the model learning through heatmaps and activation maps of individual layers of the neural network.Significance. Unexperienced users may have more time to swiftly pre-process, build/train their AI models on their own use-cases, and further visualize and explore these AI models as part of this pipeline, all in an end-to-end manner. CRP10AII will be provided as an open-source tool, and we expect to continue developing it based on users' feedback.

Interpretable Skin Cancer Classification based on Incremental Domain Knowledge Learning

The recent advances in artificial intelligence have led to the rapid development of computer-aided skin cancer diagnosis applications that perform on par with dermatologists. However, the black-box nature of such applications makes it difficult for physicians to trust the predicted decisions, subsequently preventing the proliferation of such applications in the clinical workflow. In this work, we aim to address this challenge by developing an interpretable skin cancer diagnosis approach using clinical images. Accordingly, a skin cancer diagnosis model consolidated with two interpretability methods is developed. The first interpretability method integrates skin cancer diagnosis domain knowledge, characterized by a skin lesion taxonomy, into model development, whereas the other method focuses on visualizing the decision-making process by highlighting the dominant of interest regions of skin lesion images. The proposed model is trained and validated on clinical images since the latter are easily obtainable by non-specialist healthcare providers. The results demonstrate the effectiveness of incorporating lesion taxonomy in improving model classification accuracy, where our model can predict the skin lesion origin as melanocytic or non-melanocytic with an accuracy of 87%, predict lesion malignancy with 77% accuracy, and provide disease diagnosis with an accuracy of 71%. In addition, the implemented interpretability methods assist understand the model's decision-making process and detecting misdiagnoses. This work is a step toward achieving interpretability in skin cancer diagnosis using clinical images. The developed approach can assist general practitioners to make an early diagnosis, thus reducing the redundant referrals that expert dermatologists receive for further investigations.

Deep Learning in Optical Coherence Tomography Angiography: Current Progress, Challenges, and Future Directions

Optical coherence tomography angiography (OCT-A) provides depth-resolved visualization of the retinal microvasculature without intravenous dye injection. It facilitates investigations of various retinal vascular diseases and glaucoma by assessment of qualitative and quantitative microvascular changes in the different retinal layers and radial peripapillary layer non-invasively, individually, and efficiently. Deep learning (DL), a subset of artificial intelligence (AI) based on deep neural networks, has been applied in OCT-A image analysis in recent years and achieved good performance for different tasks, such as image quality control, segmentation, and classification. DL technologies have further facilitated the potential implementation of OCT-A in eye clinics in an automated and efficient manner and enhanced its clinical values for detecting and evaluating various vascular retinopathies. Nevertheless, the deployment of this combination in real-world clinics is still in the "proof-of-concept" stage due to several limitations, such as small training sample size, lack of standardized data preprocessing, insufficient testing in external datasets, and absence of standardized results interpretation. In this review, we introduce the existing applications of DL in OCT-A, summarize the potential challenges of the clinical deployment, and discuss future research directions.

Institutional Strategies to Maintain and Grow Imaging Research During the COVID-19 Pandemic

Understanding imaging research experiences, challenges, and strategies for academic radiology departments during and after COVID-19 is critical to prepare for future disruptive events. We summarize key insights and programmatic initiatives at major academic hospitals across the world, based on literature review and meetings of the Radiological Society of North America Vice Chairs of Research (RSNA VCR) group. Through expert discussion and case studies, we provide suggested guidelines to maintain and grow radiology research in the postpandemic era.

Detecting liver cirrhosis in computed tomography scans using clinically-inspired and radiomic features

Hepatic cirrhosis is an increasing cause of mortality in developed countries-it is the pathological sequela of chronic liver diseases, and the final liver fibrosis stage. Since cirrhosis evolves from the asymptomatic phase, it is of paramount importance to detect it as quickly as possible, because entering the symptomatic phase commonly leads to hospitalization and can be fatal. Understanding the state of the liver based on the abdominal computed tomography (CT) scans is tedious, user-dependent and lacks reproducibility. We tackle these issues and propose an end-to-end and reproducible approach for detecting cirrhosis from CT. It benefits from the introduced clinically-inspired features that reflect the patient's characteristics which are often investigated by experienced radiologists during the screening process. Such features are coupled with the radiomic ones extracted from the liver, and from the suggested region of interest which captures the liver's boundary. The rigorous experiments, performed over two heterogeneous clinical datasets (two cohorts of 241 and 32 patients) revealed that extracting radiomic features from the liver's rectified contour is pivotal to enhance the classification abilities of the supervised learners. Also, capturing clinically-inspired image features significantly improved the performance of such models, and the proposed features were consistently selected as the important ones. Finally, we showed that selecting the most discriminative features leads to the Pareto-optimal models with enhanced feature-level interpretability, as the number of features was dramatically reduced (280×) from thousands to tens.

Explaining COVID-19 diagnosis with Taylor decompositions

The COVID-19 pandemic has devastated the entire globe since its first appearance at the end of 2019. Although vaccines are now in production, the number of contaminations remains high, thus increasing the number of specialized personnel that can analyze clinical exams and points out the final diagnosis. Computed tomography and X-ray images are the primary sources for computer-aided COVID-19 diagnosis, but we still lack better interpretability of such automated decision-making mechanisms. This manuscript presents an insightful comparison of three approaches based on explainable artificial intelligence (XAI) to light up interpretability in the context of COVID-19 diagnosis using deep networks: Composite Layer-wise Propagation, Single Taylor Decomposition, and Deep Taylor Decomposition. Two deep networks have been used as the backbones to assess the explanation skills of the XAI approaches mentioned above: VGG11 and VGG16. We hope that such work can be used as a basis for further research on XAI and COVID-19 diagnosis for each approach figures its own positive and negative points.

Current status and future developments in predicting outcomes in radiation oncology

Advancements in data-driven technologies and the inclusion of information-rich multiomics features have significantly improved the performance of outcomes modeling in radiation oncology. For this current trend to be sustainable, challenges related to robust data modeling such as small sample size, low size to feature ratio, noisy data, as well as issues related to algorithmic modeling such as complexity, uncertainty, and interpretability, need to be mitigated if not resolved. Emerging computational technologies and new paradigms such as federated learning, human-in-the-loop, quantum computing, and novel interpretability methods show great potential in overcoming these challenges and bridging the gap towards precision outcome modeling in radiotherapy. Examples of these promising technologies will be presented and their potential role in improving outcome modeling will be discussed.

Application of explainable artificial intelligence for healthcare: A systematic review of the last decade (2011-2022)

BACKGROUND AND OBJECTIVES

Artificial intelligence (AI) has branched out to various applications in healthcare, such as health services management, predictive medicine, clinical decision-making, and patient data and diagnostics. Although AI models have achieved human-like performance, their use is still limited because they are seen as a black box. This lack of trust remains the main reason for their low use in practice, especially in healthcare. Hence, explainable artificial intelligence (XAI) has been introduced as a technique that can provide confidence in the model's prediction by explaining how the prediction is derived, thereby encouraging the use of AI systems in healthcare. The primary goal of this review is to provide areas of healthcare that require more attention from the XAI research community.

METHODS

Multiple journal databases were thoroughly searched using PRISMA guidelines 2020. Studies that do not appear in Q1 journals, which are highly credible, were excluded.

RESULTS

In this review, we surveyed 99 Q1 articles covering the following XAI techniques: SHAP, LIME, GradCAM, LRP, Fuzzy classifier, EBM, CBR, rule-based systems, and others.

CONCLUSION

We discovered that detecting abnormalities in 1D biosignals and identifying key text in clinical notes are areas that require more attention from the XAI research community. We hope this is review will encourage the development of a holistic cloud system for a smart city.

Influence of contrast and texture based image modifications on the performance and attention shift of U-Net models for brain tissue segmentation

Contrast and texture modifications applied during training or test-time have recently shown promising results to enhance the generalization performance of deep learning segmentation methods in medical image analysis. However, a deeper understanding of this phenomenon has not been investigated. In this study, we investigated this phenomenon using a controlled experimental setting, using datasets from the Human Connectome Project and a large set of simulated MR protocols, in order to mitigate data confounders and investigate possible explanations as to why model performance changes when applying different levels of contrast and texture-based modifications. Our experiments confirm previous findings regarding the improved performance of models subjected to contrast and texture modifications employed during training and/or testing time, but further show the interplay when these operations are combined, as well as the regimes of model improvement/worsening across scanning parameters. Furthermore, our findings demonstrate a spatial attention shift phenomenon of trained models, occurring for different levels of model performance, and varying in relation to the type of applied image modification.

Improving disease classification performance and explainability of deep learning models in radiology with heatmap generators

As deep learning is widely used in the radiology field, the explainability of Artificial Intelligence (AI) models is becoming increasingly essential to gain clinicians' trust when using the models for diagnosis. In this research, three experiment sets were conducted with a U-Net architecture to improve the disease classification performance while enhancing the heatmaps corresponding to the model's focus through incorporating heatmap generators during training. All experiments used the dataset that contained chest radiographs, associated labels from one of the three conditions ["normal", "congestive heart failure (CHF)", and "pneumonia"], and numerical information regarding a radiologist's eye-gaze coordinates on the images. The paper that introduced this dataset developed a U-Net model, which was treated as the baseline model for this research, to show how the eye-gaze data can be used in multi-modal training for explainability improvement and disease classification. To compare the classification performances among this research's three experiment sets and the baseline model, the 95% confidence intervals (CI) of the area under the receiver operating characteristic curve (AUC) were measured. The best method achieved an AUC of 0.913 with a 95% CI of [0.860, 0.966]. "Pneumonia" and "CHF" classes, which the baseline model struggled the most to classify, had the greatest improvements, resulting in AUCs of 0.859 with a 95% CI of [0.732, 0.957] and 0.962 with a 95% CI of [0.933, 0.989], respectively. The decoder of the U-Net for the best-performing proposed method generated heatmaps that highlight the determining image parts in model classifications. These predicted heatmaps, which can be used for the explainability of the model, also improved to align well with the radiologist's eye-gaze data. Hence, this work showed that incorporating heatmap generators and eye-gaze information into training can simultaneously improve disease classification and provide explainable visuals that align well with how the radiologist viewed the chest radiographs when making diagnosis.

Past, Present, and Future of Machine Learning and Artificial Intelligence for Breast Cancer Screening

Breast cancer screening has evolved substantially over the past few decades because of advancements in new image acquisition systems and novel artificial intelligence (AI) algorithms. This review provides a brief overview of the history, current state, and future of AI in breast cancer screening and diagnosis along with challenges involved in the development of AI systems. Although AI has been developing for interpretation tasks associated with breast cancer screening for decades, its potential to combat the subjective nature and improve the efficiency of human image interpretation is always expanding. The rapid advancement of computational power and deep learning has increased greatly in AI research, with promising performance in detection and classification tasks across imaging modalities. Most AI systems, based on human-engineered or deep learning methods, serve as concurrent or secondary readers, that is, as aids to radiologists for a specific, well-defined task. In the future, AI may be able to perform multiple integrated tasks, making decisions at the level of or surpassing the ability of humans. Artificial intelligence may also serve as a partial primary reader to streamline ancillary tasks, triaging cases or ruling out obvious normal cases. However, before AI is used as an independent, autonomous reader, various challenges need to be addressed, including explainability and interpretability, in addition to repeatability and generalizability, to ensure that AI will provide a significant clinical benefit to breast cancer screening across all populations.

Weekly Nowcasting of New COVID-19 Cases Using Past Viral Load Measurements

The rapid spread of the coronavirus disease COVID-19 has imposed clinical and financial burdens on hospitals and governments attempting to provide patients with medical care and implement disease-controlling policies. The transmissibility of the disease was shown to be correlated with the patient’s viral load, which can be measured during testing using the cycle threshold (Ct). Previous models have utilized Ct to forecast the trajectory of the spread, which can provide valuable information to better allocate resources and change policies. However, these models combined other variables specific to medical institutions or came in the form of compartmental models that rely on epidemiological assumptions, all of which could impose prediction uncertainties. In this study, we overcome these limitations using data-driven modeling that utilizes Ct and previous number of cases, two institution-independent variables. We collected three groups of patients (n = 6296, n = 3228, and n = 12,096) from different time periods to train, validate, and independently validate the models. We used three machine learning algorithms and three deep learning algorithms that can model the temporal dynamic behavior of the number of cases. The endpoint was 7-week forward number of cases, and the prediction was evaluated using mean square error (MSE). The sequence-to-sequence model showed the best prediction during validation (MSE = 0.025), while polynomial regression (OLS) and support vector machine regression (SVR) had better performance during independent validation (MSE = 0.1596, and MSE = 0.16754, respectively), which exhibited better generalizability of the latter. The OLS and SVR models were used on a dataset from an external institution and showed promise in predicting COVID-19 incidences across institutions. These models may support clinical and logistic decision-making after prospective validation.

iCVM: An Interpretable Deep Learning Model for CVM Assessment under Label Uncertainty

The Cervical Vertebral Maturation (CVM) method aims to determine the craniofacial skeletal maturational stage, which is crucial for orthodontic and orthopedic treatment. In this paper, we explore the potential of deep learning for automatic CVM assessment. In particular, we propose a convolutional neural network named iCVM. Based on the residual network, it is specialized for the challenges unique to the task of CVM assessment. 1) To combat overfitting due to limited data size, multiple dropout layers are utilized. 2) To address the inevitable label ambiguity between adjacent maturational stages, we introduce the concept of label distribution learning in the loss function. Besides, we attempt to analyze the regions important for the prediction of the model by using the Grad-CAM technique. The learned strategy shows surprisingly high consistency with the clinical criteria. This indicates that the decisions made by our model are well interpretable, which is critical in evaluation of growth and development in orthodontics. Moreover, to drive future research in the field, we release a new dataset named CVM-900 along with the paper. It contains the cervical part of 900 lateral cephalograms collected from orthodontic patients of different ages and genders. Experimental results show that the proposed approach achieves superior performance on CVM-900 in terms of various evaluation metrics.