Search

Many real-world image recognition problems, such as diagnostic medical imaging exams, are "long-tailed" - there are a few common findings followed by many more relatively rare conditions. In chest radiography, diagnosis is both a long-tailed and multi-label problem, as patients often present with multiple findings simultaneously. While researchers have begun to study the problem of long-tailed learning in medical image recognition, few have studied the interaction of label imbalance and label co-occurrence posed by long-tailed, multi-label disease classification. To engage with the research community on this emerging topic, we conducted an open challenge, CXR-LT, on long-tailed, multi-label thorax disease classification from chest X-rays (CXRs). We publicly release a large-scale benchmark dataset of over 350,000 CXRs, each labeled with at least one of 26 clinical findings following a long-tailed distribution. We synthesize common themes of top-performing solutions, providing practical recommendations for long-tailed, multi-label medical image classification. Finally, we use these insights to propose a path forward involving vision-language foundation models for few- and zero-shot disease classification., Competing Interests: Declaration of competing interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: R.M.S has received royalties for patent or software licenses from iCAD, Philips, PingAn, ScanMed, Translation Holdings, and MGB as well as research support form CRADA with PingAn. The remaining authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Elsevier B.V. All rights reserved.)

Large chest X-rays (CXR) datasets have been collected to train deep learning models to detect thorax pathology on CXR. However, most CXR datasets are from single-center studies and the collected pathologies are often imbalanced. The aim of this study was to automatically construct a public, weakly-labeled CXR database from articles in PubMed Central Open Access (PMC-OA) and to assess model performance on CXR pathology classification by using this database as additional training data. Our framework includes text extraction, CXR pathology verification, subfigure separation, and image modality classification. We have extensively validated the utility of the automatically generated image database on thoracic disease detection tasks, including Hernia, Lung Lesion, Pneumonia, and pneumothorax. We pick these diseases due to their historically poor performance in existing datasets: the NIH-CXR dataset (112,120 CXR) and the MIMIC-CXR dataset (243,324 CXR). We find that classifiers fine-tuned with additional PMC-CXR extracted by the proposed framework consistently and significantly achieved better performance than those without (e.g., Hernia: 0.9335 vs 0.9154; Lung Lesion: 0.7394 vs. 0.7207; Pneumonia: 0.7074 vs. 0.6709; Pneumothorax 0.8185 vs. 0.7517, all in AUC with p< 0.0001) for CXR pathology detection. In contrast to previous approaches that manually submit the medical images to the repository, our framework can automatically collect figures and their accompanied figure legends. Compared to previous studies, the proposed framework improved subfigure segmentation and incorporates our advanced self-developed NLP technique for CXR pathology verification. We hope it complements existing resources and improves our ability to make biomedical image data findable, accessible, interoperable, and reusable., (Copyright © 2023 Elsevier Ltd. All rights reserved.)

Accurate identification of the presence, absence or possibility of relevant entities in clinical notes is important for healthcare professionals to quickly understand crucial clinical information. This introduces the task of assertion classification - to correctly identify the assertion status of an entity in the unstructured clinical notes. Recent rule-based and machine-learning approaches suffer from labor-intensive pattern engineering and severe class bias toward majority classes. To solve this problem, in this study, we propose a prompt-based learning approach, which treats the assertion classification task as a masked language auto-completion problem. We evaluated the model on six datasets. Our prompt-based method achieved a micro-averaged F-1 of 0.954 on the i2b2 2010 assertion dataset, with ∼1.8% improvements over previous works. In particular, our model showed excellence in detecting classes with few instances (few-shot). Evaluations on five external datasets showcase the outstanding generalizability of the prompt-based method to unseen data. To examine the rationality of our model, we further introduced two rationale faithfulness metrics: comprehensiveness and sufficiency. The results reveal that compared to the "pre-train, fine-tune" procedure, our prompt-based model has a stronger capability of identifying the comprehensive (∼63.93%) and sufficient (∼11.75%) linguistic features from free text. We further evaluated the model-agnostic explanations using LIME. The results imply a better rationale agreement between our model and human beings (∼71.93% in average F-1), which demonstrates the superior trustworthiness of our model., (Copyright © 2022 Elsevier Inc. All rights reserved.)

Rationale and Objectives: During the COVID-19 pandemic, medical educators and students are facing unprecedented challenges while navigating the new virtual landscape that social-distancing policies mandate. In response to these challenges, a new virtual introduction to radiology elective was established with unique online resources and curriculum., Materials and Methods: A previously in-person 2-week introductory radiology elective was converted into a completely virtual experience using an internally developed, open-source, peer-reviewed, web-based teaching modules combined with virtual lectures, interdisciplinary conferences, and readout sessions of de-identified cases loaded to a DICOM viewer. Students from the first four months of course enrollment completed a multiple choice pre- and post-course knowledge assessments and a 5-point Likert Scale survey as part of their educational experience., Results: In total, 26 4th-year medical students participated over 4 separate 2-week sessions from July to October of 2020. This included 12 students from the home intuition and 14 visiting students. On average, students scored 62.2% on the 55-question pre-test and 89.0% on the same test upon completion of the course, a statistically significant increase (p < 0.001). All 26 students felt engaged throughout the course. All 26 agreed (23 "strongly agreed") that they were more comfortable looking at imaging studies following the course. All 26 also agreed (21 "strongly agreed") that the course helped them prepare for their future clinical rotations and careers., Conclusion: Initial pilot program using unique web-based resources and student encounters during a two-week virtual introductory radiology elective proved to be a positive educational experience for the first 26 students enrolled., (Copyright © 2021 Elsevier Inc. All rights reserved.)

The aim of this study was to evaluate the feasibility of ultrasound strain elastography (SE) for the assessment of resting biceps brachii muscle (BBM) stiffness in patients with Parkinson's diseases (PD). From May 2014 to December 2014, we prospectively performed SE of BBM in 14 patients with PD and 10 healthy controls. Based on the Unified Parkinson's Disease Rating Scale for scoring muscle rigidity (UPDRS, part III), muscle rigidity scores in 14 patients with PD included 3 patients with high rigidity (UPDRS III-IV) and 11 patients with low rigidity (UPDRS I-II). Ultrasound strain was represented by the deformation of the BBM and subcutaneous soft tissues that was produced by external compression with a sand bag (1.5 kg) tied onto an ultrasound transducer. Deformation was estimated with two-dimensional speckle tracking. The difference in strain ratio (SR, defined as mean BBM strain divided by mean subcutaneous soft tissue strain) between PD and healthy controls was tested by unpaired t test. The correlation between SR and muscle rigidity score was analyzed by Pearson correlation coefficient. The reliability of SR in assessment of BBM stiffness was tested using intraclass correlation coefficient. In our result, the SR in PD and healthy controls measured 2.65±0.36 and 3.30±0.27, respectively. A significant difference in SR was noted between the healthy controls and PD (P=.00011). A negative correlation was found between SR and UPDRS rigidity score (r=-0.78). Our study suggests that the SR of BBM to reference tissue can be used as a quantitative biomarker in assessing resting muscle stiffness associated with muscle rigidity in PD., (Copyright © 2016 Elsevier Inc. All rights reserved.)

Mobile devices have fundamentally changed personal computing, with many people forgoing the desktop and even laptop computer altogether in favor of a smaller, lighter, and cheaper device with a touch screen. Doctors and patients are beginning to expect medical images to be available on these devices for consultative viewing, if not actual diagnosis. However, this raises serious concerns with regard to the ability of existing mobile devices and networks to quickly and securely move these images. Medical images often come in large sets, which can bog down a network if not conveyed in an intelligent manner, and downloaded data on a mobile device are highly vulnerable to a breach of patient confidentiality should that device become lost or stolen. Some degree of regulation is needed to ensure that the software used to view these images allows all relevant medical information to be visible and manipulated in a clinically acceptable manner. There also needs to be a quality control mechanism to ensure that a device's display accurately conveys the image content without loss of contrast detail. Furthermore, not all mobile displays are appropriate for all types of images. The smaller displays of smart phones, for example, are not well suited for viewing entire chest radiographs, no matter how small and numerous the pixels of the display may be. All of these factors should be taken into account when deciding where, when, and how to use mobile devices for the display of medical images., (Copyright © 2014 American College of Radiology. Published by Elsevier Inc. All rights reserved.)