Checklist for Artificial Intelligence in Medical Imaging (CLAIM): 2024 Update.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. ©RSNA, 2024.

Understanding and Mitigating Bias in Imaging Artificial Intelligence.

Artificial intelligence (AI) algorithms are prone to bias at multiple stages of model development, with potential for exacerbating health disparities. However, bias in imaging AI is a complex topic that encompasses multiple coexisting definitions. Bias may refer to unequal preference to a person or group owing to preexisting attitudes or beliefs, either intentional or unintentional. However, cognitive bias refers to systematic deviation from objective judgment due to reliance on heuristics, and statistical bias refers to differences between true and expected values, commonly manifesting as systematic error in model prediction (ie, a model with output unrepresentative of real-world conditions). Clinical decisions informed by biased models may lead to patient harm due to action on inaccurate AI results or exacerbate health inequities due to differing performance among patient populations. However, while inequitable bias can harm patients in this context, a mindful approach leveraging equitable bias can address underrepresentation of minority groups or rare diseases. Radiologists should also be aware of bias after AI deployment such as automation bias, or a tendency to agree with automated decisions despite contrary evidence. Understanding common sources of imaging AI bias and the consequences of using biased models can guide preventive measures to mitigate its impact. Accordingly, the authors focus on sources of bias at stages along the imaging machine learning life cycle, attempting to simplify potentially intimidating technical terminology for general radiologists using AI tools in practice or collaborating with data scientists and engineers for AI tool development. The authors review definitions of bias in AI, describe common sources of bias, and present recommendations to guide quality control measures to mitigate the impact of bias in imaging AI. Understanding the terms featured in this article will enable a proactive approach to identifying and mitigating bias in imaging AI. Published under a CC BY 4.0 license. Test Your Knowledge questions for this article are available in the supplemental material. See the invited commentary by Rouzrokh and Erickson in this issue.

Deep Learning for Detection of Pneumothorax and Pleural Effusion on Chest Radiographs: Validation Against Computed Tomography, Impact on Resident Reading Time, and Interreader Concordance.

PURPOSE: To study the performance of artificial intelligence (AI) for detecting pleural pathology on chest radiographs (CXRs) using computed tomography as ground truth. PATIENTS AND METHODS: Retrospective study of subjects undergoing CXR in various clinical settings. Computed tomography obtained within 24 hours of the CXR was used to volumetrically quantify pleural effusions (PEfs) and pneumothoraxes (Ptxs). CXR was evaluated by AI software (INSIGHT CXR; Lunit) and by 3 second-year radiology residents, followed by AI-assisted reassessment after a 3-month washout period. We used the area under the receiver operating characteristics curve (AUROC) to assess AI versus residents' performance and mixed-model analyses to investigate differences in reading time and interreader concordance. RESULTS: There were 96 control subjects, 165 with PEf, and 101 with Ptx. AI-AUROC was noninferior to aggregate resident-AUROC for PEf (0.82 vs 0.86, P < 0.001) and Ptx (0.80 vs 0.84, P = 0.001) detection. AI-assisted resident-AUROC was higher but not significantly different from the baseline. AI-assisted reading time was reduced by 49% (157 vs 80 s per case, P = 0.009), and Fleiss kappa for Ptx detection increased from 0.70 to 0.78 ( P = 0.003). AI decreased detection error for PEf (odds ratio = 0.74, P = 0.024) and Ptx (odds ratio = 0.39, P < 0.001). CONCLUSION: Current AI technology for the detection of PEf and Ptx on CXR was noninferior to second-year resident performance and could help decrease reading time and detection error.

Evaluation of Climate-Aware Metrics Tools for Radiology Informatics and Artificial Intelligence: Toward a Potential Radiology Ecolabel.

Radiology is a major contributor to health care's impact on climate change, in part due to its reliance on energy-intensive equipment as well as its growing technological reliance. Delivering modern patient care requires a robust informatics team to move images from the imaging equipment to the workstations and the health care system. Radiology informatics is the field that manages medical imaging IT. This involves the acquisition, storage, retrieval, and use of imaging information in health care to improve access and quality, which includes PACS, cloud services, and artificial intelligence. However, the electricity consumption of computing and the life cycle of various computer components expands the carbon footprint of health care. The authors provide a general framework to understand the environmental impact of clinical radiology informatics, which includes using the international Greenhouse Gas Protocol to draft a definition of scopes of emissions pertinent to radiology informatics, as well as exploring existing tools to measure and account for these emissions. A novel standard ecolabel for radiology informatics tools, such as the Energy Star label for consumer devices or Leadership in Energy and Environmental Design certification for buildings, should be developed to promote awareness and guide radiologists and radiology informatics leaders in making environmentally conscious decisions for their clinical practice. At this critical climate juncture, the radiology community has a unique and pressing obligation to consider our shared environmental responsibility in innovating clinical technology for patient care.

Artificial Intelligence and Radiology Education.

Implementation of artificial intelligence (AI) applications into clinical practice requires AI-savvy radiologists to ensure the safe, ethical, and effective use of these systems for patient care. Increasing demand for AI education reflects recognition of the translation of AI applications from research to clinical practice, with positive trainee attitudes regarding the influence of AI on radiology. However, barriers to AI education, such as limited access to resources, predispose to insufficient preparation for the effective use of AI in practice. In response, national organizations have sponsored formal and self-directed learning courses to provide introductory content on imaging informatics and AI. Foundational courses, such as the National Imaging Informatics Course - Radiology and the Radiological Society of North America Imaging AI Certificate, lay a framework for trainees to explore the creation, deployment, and critical evaluation of AI applications. This report includes additional resources for formal programming courses, video series from leading organizations, and blogs from AI and informatics communities. Furthermore, the scope of "AI and radiology education" includes AI-augmented radiology education, with emphasis on the potential for "precision education" that creates personalized experiences for trainees by accounting for varying learning styles and inconsistent, possibly deficient, clinical case volume. © RSNA, 2022 Keywords: Use of AI in Education, Impact of AI on Education, Artificial Intelligence, Medical Education, Imaging Informatics, Natural Language Processing, Precision Education.

American Society of Emergency Radiology (ASER) social media committee workgroup: best practices for the use of social media in emergency radiology.

Social media has become integrated within the profession of medicine, and emergency radiology has inevitably felt the impact of its presence. Emergency radiologists are encouraged to consider the advantages of embracing the digital era and the benefits it may bring to our careers. We aim to present the best practice guidelines for emergency radiologists and radiology departments. This paper is a product of the American Society of Emergency Radiology Social Media (ASER) Committee workgroup and represents the best practices of the society.

Performance of Multiple Pretrained BERT Models to Automate and Accelerate Data Annotation for Large Datasets.

Purpose: To develop and evaluate domain-specific and pretrained bidirectional encoder representations from transformers (BERT) models in a transfer learning task on varying training dataset sizes to annotate a larger overall dataset. Materials and Methods: The authors retrospectively reviewed 69 095 anonymized adult chest radiograph reports (reports dated April 2020-March 2021). From the overall cohort, 1004 reports were randomly selected and labeled for the presence or absence of each of the following devices: endotracheal tube (ETT), enterogastric tube (NGT, or Dobhoff tube), central venous catheter (CVC), and Swan-Ganz catheter (SGC). Pretrained transformer models (BERT, PubMedBERT, DistilBERT, RoBERTa, and DeBERTa) were trained, validated, and tested on 60%, 20%, and 20%, respectively, of these reports through fivefold cross-validation. Additional training involved varying dataset sizes with 5%, 10%, 15%, 20%, and 40% of the 1004 reports. The best-performing epochs were used to assess area under the receiver operating characteristic curve (AUC) and determine run time on the overall dataset. Results: The highest average AUCs from fivefold cross-validation were 0.996 for ETT (RoBERTa), 0.994 for NGT (RoBERTa), 0.991 for CVC (PubMedBERT), and 0.98 for SGC (PubMedBERT). DeBERTa demonstrated the highest AUC for each support device trained on 5% of the training set. PubMedBERT showed a higher AUC with a decreasing training set size compared with BERT. Training and validation time was shortest for DistilBERT at 3 minutes 39 seconds on the annotated cohort. Conclusion: Pretrained and domain-specific transformer models required small training datasets and short training times to create a highly accurate final model that expedites autonomous annotation of large datasets.Keywords: Informatics, Named Entity Recognition, Transfer Learning Supplemental material is available for this article. ©RSNA, 2022See also the commentary by Zech in this issue.

Incidental Intercostal Intramuscular Myxoma With Fibrous Dysplasia in a Patient With Mazabraud's and McCune Albright Syndromes.

Intramuscular myxoma is a rare entity that may present as single or multiple lesions in patients with Mazabraud's syndrome and is characterized by intramuscular myxomas with fibrous dysplasia. Though intramuscular myxomas occur in large muscle groups, they can very rarely occur in the chest wall. We present the case of a 41-year-old woman with an incidentally discovered intercostal mass on magnetic resonance cholangiopancreatography (MRCP). Repeat MRI demonstrated a lobulated, T2-hyperintense intercostal lesion and demonstrated adjacent fibrous dysplasia of the ribs, consistent with the patient's history of Mazabraud's and McCune Albright syndromes. Histopathological exam following surgical resection confirmed a diagnosis of intramuscular myxoma without the presence of sarcomatous changes. Though small, slow-growing intramuscular myxomas may be observed with conservative management in the absence of significant symptoms, surgical resection is warranted to prevent complications such as osseous erosion or nerve impingement.

Concurrent, Bilateral Presentation of Immature and Mature Ovarian Teratomas with Refractory Hyponatremia: A Case Report.

We present the imaging and histopathological findings in a 32-year-old female who presented to the erectile dysfunction with progressively worsening abdominal pain over the past 2 months. Computed tomography abdomen and pelvis revealed bilateral ovarian teratomas, left significantly larger than right. There was associated fat stranding, mesenteric/omental stranding, and ascites worrisome for rupture versus peritoneal carcinomatosis. Histopathology confirmed a left immature teratoma (Grade 2), right mature teratoma, and peritoneal gliomatosis from possible tumor rupture before surgery.

IMPLEMENTING A STANDARDIZED INTERVENTIONAL EXERCISE REGIMEN TO IMPROVE FUNCTIONAL MOVEMENTS IN FEMALE COLLEGIATE ATHLETES.

BACKGROUND: Interventional exercises have been developed to help athletes improve scores on the Functional Movement Screen™ (FMS™). However, there is a paucity of research on the effects of a similar program in female athletes, as well as the effects of a standardized corrective exercise regimen. The purpose of this study was to assess whether an in-season, standardized interventional exercise program improves FMS™ score asymmetry and the composite score of female collegiate athletes. STUDY DESIGN: Prospective, quasi-experimental, cohort study. METHODS: Forty-one (mean age 19.5 ± 1.2 years; body mass, 70.6 ± 11.5 kg ; height, 1.70 ± 0.083 m) NCAA Division III female soccer (n=10), softball (n=17), and basketball (n=14) players participated in this study. The athletes completed the FMS™ screens prior to their season, regularly participated in four in-season standardized corrective exercises throughout three to four month athletic seasons, and completed the FMS™ screens in the postseason. RESULTS: The average score of all athletes before the season was 15.52 ± 0.63 and 16.04 ± 0.72 after the season. While the mean score of soccer players increased from 14.80 ± 0.92 to 16.1 ± 1.52 and the mean score of softball players increased from 15.83 ± 1.89 to 16.72 ± 1.41 at the end of the season, the mean score of basketball players dropped from 15.93 ± 1.49 to 15.29 ± 1.59. Women's basketball players experienced a decrease in their composite FMS™ score ( x ¯ = -0.571, p<0.01), while women's soccer players ( x ¯ =+1.30, p<0.05) and softball players ( x ¯ =+1.12, p<0.05) experienced an increase in mean score 2.28 times and 1.96 times greater in magnitude than the decrease in basketball players' composite FMS™, respectively. Fewer total athletes demonstrated asymmetries at postseason testing, decreasing from 24 at preseason testing to 15 at postseason testing (p<0.01). Significant differences were not noted between athlete age and FMS™ scores (p>0.05). CONCLUSIONS: Standardized interventional programs during athletic teams' seasons may be used to help increase FMS™ scores and reduce asymmetry. Though more studies are warranted to address the negative effects of this standardized program in women's basketball players, this study demonstrated that the number of asymmetries significantly decreased from pre- to postseason among soccer and softball players, which may have implications for a higher resistance to injury. LEVELS OF EVIDENCE: 3.

Orally active opioid compounds from a non-poppy source.

The basic science and clinical use of morphine and other "opioid" drugs are based almost exclusively on the extracts or analogues of compounds isolated from a single source, the opium poppy (Papaver somniferum). However, it now appears that biological diversity has evolved an alternative source. Specifically, at least two alkaloids isolated from the plant Mitragyna speciosa, mitragynine ((E)-2-[(2S,3S)-3-ethyl-8-methoxy-1,2,3,4,6,7,12,12b-octahydroindolo[3,2-h]quinolizin-2-yl]-3-methoxyprop-2-enoic acid methyl ester; 9-methoxy coryantheidine; MG) and 7-hydroxymitragynine (7-OH-MG), and several synthetic analogues of these natural products display centrally mediated (supraspinal and spinal) antinociceptive (analgesic) activity in various pain models. Several characteristics of these compounds suggest a classic "opioid" mechanism of action: nanomolar affinity for opioid receptors, competitive interaction with the opioid receptor antagonist naloxone, and two-way analgesic cross-tolerance with morphine. However, other characteristics of the compounds suggest novelty, particularly chemical structure and possible greater separation from side effects. We review the chemical and pharmacological properties of these compounds.