Evaluating GPT-V4 (GPT-4 with Vision) on Detection of Radiologic Findings on Chest Radiographs.

Background Generating radiologic findings from chest radiographs is pivotal in medical image analysis. The emergence of OpenAI's generative pretrained transformer, GPT-4 with vision (GPT-4V), has opened new perspectives on the potential for automated image-text pair generation. However, the application of GPT-4V to real-world chest radiography is yet to be thoroughly examined. Purpose To investigate the capability of GPT-4V to generate radiologic findings from real-world chest radiographs. Materials and Methods In this retrospective study, 100 chest radiographs with free-text radiology reports were annotated by a cohort of radiologists, two attending physicians and three residents, to establish a reference standard. Of 100 chest radiographs, 50 were randomly selected from the National Institutes of Health (NIH) chest radiographic data set, and 50 were randomly selected from the Medical Imaging and Data Resource Center (MIDRC). The performance of GPT-4V at detecting imaging findings from each chest radiograph was assessed in the zero-shot setting (where it operates without prior examples) and few-shot setting (where it operates with two examples). Its outcomes were compared with the reference standard with regards to clinical conditions and their corresponding codes in the International Statistical Classification of Diseases, Tenth Revision (ICD-10), including the anatomic location (hereafter, laterality). Results In the zero-shot setting, in the task of detecting ICD-10 codes alone, GPT-4V attained an average positive predictive value (PPV) of 12.3%, average true-positive rate (TPR) of 5.8%, and average F1 score of 7.3% on the NIH data set, and an average PPV of 25.0%, average TPR of 16.8%, and average F1 score of 18.2% on the MIDRC data set. When both the ICD-10 codes and their corresponding laterality were considered, GPT-4V produced an average PPV of 7.8%, average TPR of 3.5%, and average F1 score of 4.5% on the NIH data set, and an average PPV of 10.9%, average TPR of 4.9%, and average F1 score of 6.4% on the MIDRC data set. With few-shot learning, GPT-4V showed improved performance on both data sets. When contrasting zero-shot and few-shot learning, there were improved average TPRs and F1 scores in the few-shot setting, but there was not a substantial increase in the average PPV. Conclusion Although GPT-4V has shown promise in understanding natural images, it had limited effectiveness in interpreting real-world chest radiographs. © RSNA, 2024 Supplemental material is available for this article.

Measurement of adherence and health-related quality of life during anticoagulation therapy in cancer-associated venous thromboembolism (VTE): a multicenter quantitative study.

PURPOSE: Therapy for cancer-associated venous thromboembolism (VTE) includes long-term anticoagulation, which may have substantial impact on the health-related quality of life (HRQL) of patients. We assessed patient-reported outcomes to characterize the HRQL associated with VTE treatment and to begin to examine those HRQL elements impacting anticoagulation adherence (AA). METHODS: Participants were adult cancer patients with confirmed symptomatic acute lower extremity deep venous thrombosis. Patients were excluded if there was an indication for anticoagulation other than VTE, ECOG performance status >3, or life expectancy < 3 months. Participants were assessed with a self-reported adherence tool. HRQL was measured with a 6-domain questionnaire using a seven-point Likert scale. Evaluations were performed at 30 days and 3 months after enrollment. For the primary objective, an overall adherence rate was calculated at each time point of evaluation. For the HRQL domains, non-parametric testing was used to compare results between subgroups. RESULTS: Seventy-four patients were enrolled. AA and HRQL at 30 days and 3 months were assessed in 50 and 36 participants, respectively. At 30 days the AA rate was 90%, and at 3 months it was 83%. In regard to HRQL, patients suffered frequent and moderate-severe distress in the domains of emotional and physical symptoms, sleep disturbance, and limitations to physical activity. An association between emotional or physical distress and AA was observed. CONCLUSION: Patients with VTE suffer a substantial impairment of their HRQL. Increased emotional distress correlated with better long-term AA. These results can be used to inform additional research aimed at developing novel strategies to improve AA.

Solid Attenuation Components Attention Deep Learning Model to Predict Micropapillary and Solid Patterns in Lung Adenocarcinomas on Computed Tomography.

BACKGROUND: High-grade adenocarcinoma subtypes (micropapillary and solid) treated with sublobar resection have an unfavorable prognosis compared with those treated with lobectomy. We investigated the potential of incorporating solid attenuation component masks with deep learning in the prediction of high-grade components to optimize surgical strategy preoperatively. METHODS: A total of 502 patients with pathologically confirmed high-grade adenocarcinomas were retrospectively enrolled between 2016 and 2020. The SACs attention DL model was developed to apply solid-attenuation-component-like subregion masks (tumor area ≥ - 190 HU) to guide the DL model for predicting high-grade subtypes. The SACA-DL was assessed using 5-fold cross-validation and external validation in the training and testing sets, respectively. The performance, which was evaluated using the area under the curve (AUC), was compared between SACA-DL and the DL model without SACs attention (DLwoSACs), the prior radiomics model, or the model based on the consolidation/tumor (C/T) diameter ratio. RESULTS: We classified 313 and 189 patients into training and testing cohorts, respectively. The SACA-DL achieved an AUC of 0.91 for the cross-validation, which was significantly superior to those of the DLwoSACs (AUC = 0.88; P = 0.02), prior radiomics model (AUC = 0.85; P = 0.004), and C/T ratio (AUC = 0.84; P = 0.002). An AUC of 0.93 was achieved for external validation in the SACA-DL and was significantly better than those of the DLwoSACs (AUC = 0.89; P = 0.04), prior radiomics model (AUC = 0.85; P < 0.001), and C/T ratio (AUC = 0.85; P < 0.001). CONCLUSIONS: The combination of solid-attenuation-component-like subregion masks with the DL model is a promising approach for the preoperative prediction of high-grade adenocarcinoma subtypes.

Imaging of the mediastinum: Mimics of malignancy.

In the imaging of the mediastinum, benign lesions mimicking malignancy constitute potential pitfalls in interpretation. Localization and characteristic imaging features are key to narrow the differential diagnosis and avoid potential pitfalls in interpretation. Based on certain anatomic landmarks, the mediastinal compartment model enables accurate localization. Depending on the anatomic origin, mediastinal lesions can have various etiologies. The anatomic location and structures contained within each mediastinal compartment are helpful in generating the differential diagnoses. These structures include thyroid, thymus, parathyroid, lymph nodes, pericardium, embryogenic remnants, and parts of the enteric tracts, vessels, and nerves. Imaging characteristics on computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography-computed tomography (PET/CT), including attenuation (fluid, fat, calcification), contrast enhancement, and metabolic activity, aid in narrowing the differential diagnoses. Understanding the roles and limitations of various imaging modalities is helpful in the evaluation of mediastinal masses. In this review, we present potential pitfalls in the imaging of mediastinal lesions with emphasis on the mimics of malignancy.

Stenophyllols A-C, new compounds from Boesenbergia stenophylla.

Three new compounds, i.e. stenophyllols A-C (1-3), were isolated from the rhizome of Boesenbergia stenophylla. The structures were determined by spectroscopic analysis (UV, IR, NMR and HRESIMS). In-vitro neuroblastoma cell viability assay showed stenophyllol A (1) was able to reduce the N2A cell viability to 20% within 24 h.

The RSNA International COVID-19 Open Radiology Database (RICORD).

The coronavirus disease 2019 (COVID-19) pandemic is a global health care emergency. Although reverse-transcription polymerase chain reaction testing is the reference standard method to identify patients with COVID-19 infection, chest radiography and CT play a vital role in the detection and management of these patients. Prediction models for COVID-19 imaging are rapidly being developed to support medical decision making. However, inadequate availability of a diverse annotated data set has limited the performance and generalizability of existing models. To address this unmet need, the RSNA and Society of Thoracic Radiology collaborated to develop the RSNA International COVID-19 Open Radiology Database (RICORD). This database is the first multi-institutional, multinational, expert-annotated COVID-19 imaging data set. It is made freely available to the machine learning community as a research and educational resource for COVID-19 chest imaging. Pixel-level volumetric segmentation with clinical annotations was performed by thoracic radiology subspecialists for all COVID-19-positive thoracic CT scans. The labeling schema was coordinated with other international consensus panels and COVID-19 data annotation efforts, the European Society of Medical Imaging Informatics, the American College of Radiology, and the American Association of Physicists in Medicine. Study-level COVID-19 classification labels for chest radiographs were annotated by three radiologists, with majority vote adjudication by board-certified radiologists. RICORD consists of 240 thoracic CT scans and 1000 chest radiographs contributed from four international sites. It is anticipated that RICORD will ideally lead to prediction models that can demonstrate sustained performance across populations and health care systems.

Geometric and dosimetric accuracy of deformable image registration between average-intensity images for 4DCT-based adaptive radiotherapy for non-small cell lung cancer.

PURPOSE: Re-planning for four-dimensional computed tomography (4DCT)-based lung adaptive radiotherapy commonly requires deformable dose mapping between the planning average-intensity image (AVG) and the newly acquired AVG. However, such AVG-AVG deformable image registration (DIR) lacks accuracy assessment. The current work quantified and compared geometric accuracies of AVG-AVG DIR and corresponding phase-phase DIRs, and subsequently investigated the clinical impact of such AVG-AVG DIR on deformable dose mapping. METHODS AND MATERIALS: Hybrid intensity-based AVG-AVG and phase-phase DIRs were performed between the planning and mid-treatment 4DCTs of 28 non-small cell lung cancer patients. An automated landmark identification algorithm detected vessel bifurcation pairs in both lungs. Target registration error (TRE) of these landmark pairs was calculated for both DIR types. The correlation between TRE and respiratory-induced landmark motion in the planning 4DCT was analyzed. Global and local dose metrics were used to assess the clinical implications of AVG-AVG deformable dose mapping with both DIR types. RESULTS: TRE of AVG-AVG and phase-phase DIRs averaged 3.2 ± 1.0 and 2.6 ± 0.8 mm respectively (p < 0.001). Using AVG-AVG DIR, TREs for landmarks with <10 mm motion averaged 2.9 ± 2.0 mm, compared to 3.1 ± 1.9 mm for the remaining landmarks (p < 0.01). Comparatively, no significant difference was demonstrated for phase-phase DIRs. Dosimetrically, no significant difference in global dose metrics was observed between doses mapped with AVG-AVG DIR and the phase-phase DIR, but a positive linear relationship existed (p = 0.04) between the TRE of AVG-AVG DIR and local dose difference. CONCLUSIONS: When the region of interest experiences <10 mm respiratory-induced motion, AVG-AVG DIR may provide sufficient geometric accuracy; conversely, extra attention is warranted, and phase-phase DIR is recommended. Dosimetrically, the differences in geometric accuracy between AVG-AVG and phase-phase DIRs did not impact global lung-based metrics. However, as more localized dose metrics are needed for toxicity assessment, phase-phase DIR may be required as its lower mean TRE improved voxel-based dosimetry.

Total Lesion Glycolysis Assessment Identifies a Patient Fraction With a High Cure Rate Among Esophageal Adenocarcinoma Patients Treated With Definitive Chemoradiation.

OBJECTIVE: We aimed to determine whether tumor metabolism could be prognostic of cure in L-EAC patients who receive definitive chemoradiation. SUMMARY BACKGROUND DATA: Patients with inoperable localized esophageal adenocarcinoma (L-EAC) often receive definitive chemoradiation; however, biomarkers and/or imaging variables to prognosticate cure are missing. METHODS: Two hundred sixty-six patients with L-EAC who had chemoradiation but not surgery were analyzed from the prospectively maintained EAC databases in the Department of Gastrointestinal Medical Oncology at The University of Texas MD Anderson Cancer Center (Texas, USA) between March 2002 and April 2015. Maximum standardized uptake value (SUVmax) and total lesion glycolysis (TLG) from the positron emission tomography data were evaluated. RESULTS: Of 266 patients, 253 (95%) were men; the median age was 67 years (range 20-91 yrs) and 153 had poorly differentiated L-EAC. The median SUVmax was 10.3 (range 0-87) and the median TLG was 85.7 (range 0-3227). Both SUVmax and TLG were higher among those with: tumors >5 cm in length, high clinical stage, and high tumor and node categories by TNM staging (all P < 0.0001). Of 234 patients evaluable for cure, 60 (25.6%) achieved cure. In the multivariable logistic regression model, low TLG (but not low SUVmax) was associated with cure (continuous TLG value: odds ratio 0.70, 95% confidence interval (CI) 0.54-0.92). TLG was quantified into 4 quartile categorical variables; first quartile (Q1; <32), second quartile (Q2; 32.0-85.6), third quartile (Q3; 85.6-228.4), and fourth quartile (Q4; >228.4); the cure rate was only 10.3% in Q4 and 5.1% in Q3 but increased to 28.8% in Q2, and 58.6% in Q1. The cross-validation resulted in an average accuracy of prediction score of 0.81 (95% CI, 0.75-0.86). CONCLUSIONS: In this cross-validated model, 59% of patients in the 1st quartile were cured following definitive chemoradiation. Baseline TLG could be pursued as one of the tools for esophageal preservation.

Evolution of CT findings in patients with mild COVID-19 pneumonia.

OBJECTIVES: To delineate the evolution of CT findings in patients with mild COVID-19 pneumonia. METHODS: CT images and medical records of 88 patients with confirmed mild COVID-19 pneumonia, a baseline CT, and at least one follow-up CT were retrospectively reviewed. CT features including lobar distribution and presence of ground glass opacities (GGO), consolidation, and linear opacities were analyzed on per-patient basis during each of five time intervals spanning the 3 weeks after disease onset. Total severity scores were calculated. RESULTS: Of patients, 85.2% had travel history to Wuhan or known contact with infected individuals. The most common symptoms were fever (84.1%) and cough (56.8%). The baseline CT was obtained on average 5 days from symptom onset. Four patients (4.5%) had negative initial CT. Significant differences were found among the time intervals in the proportion of pulmonary lesions that are (1) pure GGO, (2) mixed attenuation, (3) mixed attenuation with linear opacities, (4) consolidation with linear opacities, and (5) pure consolidation. The majority of patients had involvement of ≥ 3 lobes. Bilateral involvement was more prevalent than unilateral involvement. The proportions of patients observed to have pure GGO or GGO and consolidation decreased over time while the proportion of patients with GGO and linear opacities increased. Total severity score showed an increasing trend in the first 2 weeks. CONCLUSIONS: While bilateral GGO are predominant features, CT findings changed during different time intervals in the 3 weeks after symptom onset in patients with COVID-19. KEY POINTS: • Four of 88 (4.5%) patients with COVID-19 had negative initial CT. • Majority of COVID-19 patients had abnormal CT findings in ≥ 3 lobes. • A proportion of patients with pure ground glass opacities decreased over the 3 weeks after symptom onset.

In Situ Pulmonary Artery Thrombosis: Unrecognized Complication of Radiation Therapy.

OBJECTIVE. The purpose of this study is to evaluate the CT and clinical characteristics of in situ pulmonary artery thrombosis (PAT) associated with radiation therapy (RT). MATERIALS AND METHODS. A database search was performed to identify patients who had PAT develop after receiving RT. The CT characteristics of PAT, including the number, location, and appearance of filling defects as well as the presence of associated lung fibrosis, were recorded. The terminology (in situ thrombosis vs acute or chronic pulmonary embolism) used by the interpreting radiologists to describe PAT, the time between the completion of RT and development of PAT, the change in the size of the PAT, and observation of any new thrombi and emboli on follow-up imaging, were also recorded. RESULTS. Of the 27 patients in the study cohort, 22 (81%) had lung cancer and five (19%) had mesothelioma. Most PATs were solitary (93%) and nonocclusive (96%) and formed an obtuse angle to the vessel wall (89%). All PATs were eccentric within the involved PA and were located within the RT volume. The time from completion of RT to initial diagnosis of PAT on CT ranged from 53 to 2522 days (mean, 675 days). Radiation-induced lung fibrosis was present in the ipsilateral lung in all patients. No evidence of additional PA filling defects that suggested embolization were seen on follow-up images of any of the patients, even those who did not receive anticoagulant therapy. CONCLUSION. In situ PAT associated with RT, which to our knowledge has not previously been described in the English literature, has imaging features different from those of acute pulmonary emboli and does not appear to embolize. Radiologist awareness of PAT can facilitate accurate diagnosis and impact management.

Clinical and Cancer-Related Predictors for Venous Thromboembolism in Cancer Patients Presenting to the Emergency Department.

BACKGROUND: The accurate detection of cancer-associated venous thromboembolism (VTE) can avoid unnecessary diagnostic imaging or laboratory tests. OBJECTIVE: We sought to determine clinical and cancer-related risk factors of VTE that can be used as predictors for oncology patients presenting to the emergency department (ED) with suspected VTE. METHODS: We retrospectively analyzed all consecutive patients who presented with suspicion of VTE to The University of Texas MD Anderson Cancer Center ED between January 1, 2009, and January 1, 2013. Logistic regression models were used to identify risk factors that were associated with VTE. The ability of these factors to predict VTE was externally validated using a second cohort of patients who presented to King Hussein Cancer Center ED between January 1, 2009, and January 1, 2016. RESULTS: Cancer-related covariates associated with the occurrence of VTE were high-risk cancer type (odds ratio [OR] 3.64 [95% confidence interval {CI} 2.37-5.60], p < 0.001), presentation within 6 months of the cancer diagnosis (OR 1.92 [95% CI 1.62-2.28], p < 0.001), active cancer (OR 1.35 [95% CI 1.10-1.65], p = 0.003), advanced stage (OR 1.40 [95% CI 1.01-1.94], p = 0.044), and the presence of brain metastasis (OR 1.73 [95% CI 1.32-2.27], p < 0.001). When combined, these factors along with other clinical factors showed high prediction performance for VTE in the external validation cohort. CONCLUSIONS: Cancer risk group, presentation within 6 months of cancer diagnosis, active and advanced cancer, and the presence of brain metastases along with other related clinical factors can be used to predict VTE in patients with cancer presenting to the ED.

Crowdsourcing pneumothorax annotations using machine learning annotations on the NIH chest X-ray dataset.

Pneumothorax is a potentially life-threatening condition that requires prompt recognition and often urgent intervention. In the ICU setting, large numbers of chest radiographs are performed and must be interpreted on a daily basis which may delay diagnosis of this entity. Development of artificial intelligence (AI) techniques to detect pneumothorax could help expedite detection as well as localize and potentially quantify pneumothorax. Open image analysis competitions are useful in advancing state-of-the art AI algorithms but generally require large expert annotated datasets. We have annotated and adjudicated a large dataset of chest radiographs to be made public with the goal of sparking innovation in this space. Because of the cumbersome and time-consuming nature of image labeling, we explored the value of using AI models to generate annotations for review. Utilization of this machine learning annotation (MLA) technique appeared to expedite our annotation process with relatively high sensitivity at the expense of specificity. Further research is required to confirm and better characterize the value of MLAs. Our adjudicated dataset is now available for public consumption in the form of a challenge.

Multi-institutional Analysis of Recurrence and Survival After Neoadjuvant Chemoradiotherapy of Esophageal Cancer: Impact of Histology on Recurrence Patterns and Outcomes.

OBJECTIVE: To determine the impact of histology on pathologic response, survival outcomes, and recurrence patterns in patients with esophageal cancer (EC) who received neoadjuvant chemoradiotherapy (CRT). SUMMARY OF BACKGROUND DATA: There is a paucity of data regarding comparative outcomes after neoadjuvant CRT between esophageal squamous cell carcinoma (SCC) and adenocarcinoma. METHODS: Between 2002 and 2015, 895 EC patients who underwent neoadjuvant CRT followed by esophagectomy at 3 academic institutions were retrospectively reviewed, including 207 patients with SCC (23.1%) and 688 patients with adenocarcinoma (76.9%). Pathologic response, survival, recurrence pattern, and potential prognostic factors were compared. RESULTS: Pathologic complete response (pCR) rate was significantly higher for SCC compared with adenocarcinoma (44.9% vs 25.9%, P < 0.001). After a median follow-up of 52.9 months, 71 patients (34.3%) with SCC versus 297 patients (43.2%) with adenocarcinoma had recurrent disease (P = 0.023). For patients who achieved a pCR, no significant differences were found in recurrence pattern, sites, or survival end-points between the 2 histology groups. For non-pCR patients, the SCC group demonstrated significantly higher regional and supraclavicular recurrence rates but a lower hematogenous metastasis rate than adenocarcinoma patients, whereas the adenocarcinoma patients had a more favorable locoregional failure-free survival (P = 0.005) and worse distant metastasis-free survival (P = 0.024). No differences were found in overall survival (P = 0.772) or recurrence-free survival (P = 0.696) between groups. CONCLUSIONS: SCC was associated with a significantly higher pCR rate than adenocarcinoma. Recurrence pattern and survival outcomes were significantly different between the 2 histology subtypes in non-pCR patients.

Ethics of Artificial Intelligence in Radiology: Summary of the Joint European and North American Multisociety Statement.

This is a condensed summary of an international multisociety statement on ethics of artificial intelligence (AI) in radiology produced by the ACR, European Society of Radiology, RSNA, Society for Imaging Informatics in Medicine, European Society of Medical Imaging Informatics, Canadian Association of Radiologists, and American Association of Physicists in Medicine. AI has great potential to increase efficiency and accuracy throughout radiology, but it also carries inherent pitfalls and biases. Widespread use of AI-based intelligent and autonomous systems in radiology can increase the risk of systemic errors with high consequence and highlights complex ethical and societal issues. Currently, there is little experience using AI for patient care in diverse clinical settings. Extensive research is needed to understand how to best deploy AI in clinical practice. This statement highlights our consensus that ethical use of AI in radiology should promote well-being, minimize harm, and ensure that the benefits and harms are distributed among stakeholders in a just manner. We believe AI should respect human rights and freedoms, including dignity and privacy. It should be designed for maximum transparency and dependability. Ultimate responsibility and accountability for AI remains with its human designers and operators for the foreseeable future. The radiology community should start now to develop codes of ethics and practice for AI that promote any use that helps patients and the common good and should block use of radiology data and algorithms for financial gain without those two attributes. This article is a simultaneous joint publication in Radiology, Journal of the American College of Radiology, Canadian Association of Radiologists Journal, and Insights into Imaging. Published under a CC BY-NC-ND 4.0 license. Online supplemental material is available for this article.

Leukemic Involvement in the Thorax.

Leukemias are malignancies in which abnormal white blood cells are produced in the bone marrow, resulting in compromise of normal bone marrow hematopoiesis and subsequent cytopenias. Leukemias are classified as myeloid or lymphoid depending on the type of abnormal cells produced and as acute or chronic according to cellular maturity. The four major types of leukemia are acute myeloid leukemia, chronic myeloid leukemia, acute lymphoblastic leukemia, and chronic lymphocytic leukemia. Clinical manifestations are due to either bone marrow suppression (anemia, thrombocytopenia, or neutropenia) or leukemic organ infiltration. Imaging manifestations of leukemia in the thorax are myriad. While lymphadenopathy is the most common manifestation of intrathoracic leukemia, leukemia may also involve the lungs, pleura, heart, and bones and soft tissues. Myeloid sarcomas occur in 5%-7% of patients with acute myeloid leukemia and represent masses of myeloid blast cells in an extramedullary location. ©RSNA, 2019.

Imaging Findings in Transgender Patients after Gender-affirming Surgery.

Gender-affirming surgeries expand the options for physical transition among transgender patients, those whose gender identity is incongruent with the sex assigned to them at birth. Growing medical insight, increasing public acceptance, and expanding insurance coverage have improved the access to and increased the demand for gender-affirming surgeries in the United States. Procedures for transgender women, those patients with feminine gender identity, include breast augmentation using implants and genital reconstruction with vaginoplasty. Some transgender women receive medically unapproved silicone injections for breast augmentation or other soft-tissue contouring procedures that can lead to disfigurement, silicone pulmonary embolism, systemic reactions, and even death. MRI is preferred over CT for postvaginoplasty evaluation given its superior tissue contrast resolution. Procedures for transgender men, patients with a masculine gender identity, include chest masculinization (mastectomy) and genital reconstruction (phalloplasty or metoidioplasty, scrotoplasty, and erectile device implantation). Urethrography is the standard imaging modality performed to evaluate neourethral patency and other complications, such as leaks and fistulas. Despite a sizeable growth in the surgical literature about gender-affirming surgeries and their outcomes, detailed descriptions of the imaging features following these surgeries remain sparse. Radiologists must be aware of the wide variety of anatomic and pathologic changes unique to patients who undergo gender-affirming surgeries to ensure accurate imaging interpretation. Online supplemental material is available for this article. ©RSNA, 2019.

Normal D-dimer levels in cancer patients with radiologic evidence of pulmonary embolism.

Accurate and expeditious diagnosis and treatment of pulmonary embolism in cancer patients improves patient outcomes. D-dimer is often used to rule out pulmonary embolism. However, this test is less accurate in cancer patients, and it is unclear whether cancer patients with normal D-dimer levels can present with pulmonary embolism. All consecutive patients who presented to The University of Texas MD Anderson Cancer Center in Houston, Texas, USA, between May 2009 and November 2015 who underwent computed tomography pulmonary angiography and plasma D-dimer level measurement were retrospectively reviewed. Patients with suspected pulmonary embolism and normal D-dimer levels were identified. Among the 8023 cancer patients identified, 1156 (14%) had pulmonary embolism. Only 35 patients with pulmonary embolism (3%) had normal plasma D-dimer levels. Twenty-six of these patients had acute pulmonary embolism and the other nine had subacute or chronic pulmonary embolism. Thirteen of the 26 acute cases were in patients with hematological cancer. Most patients (23/35, 66%) had subsegmental or segmental pulmonary embolism. Only one patient had pulmonary embolism in the main pulmonary arteries. Although it is uncommon (3%), cancer patients with radiologic evidence of pulmonary embolism can present with normal D-dimer levels. Recognizing the possibility of this uncommon occurrence is critical in the decision process for ordering diagnostic tests for evaluation of suspected pulmonary embolism.

Natural History of Persistent Pulmonary Subsolid Nodules: Long-Term Observation of Different Interval Growth.

BACKGROUND: The long-term natural course and outcomes of subsolid nodules (SSNs) in terms of true growth, substantial growth, and stage shift need to be clarified. METHODS: Between 2002 and 2016, 128 subjects with persistent SSNs of 3cm or smaller were enrolled. The baseline and interval changes in the series computed tomography (CT) findings during the follow-up period were subsequently reviewed. RESULTS: The mean follow-up period was 3.57±2.93years. The cumulative percentage of growth nodules of the part-solid nodule (PSN) group was significantly higher than that of the ground-glass nodule (GGN) group by Kaplan-Meier estimation (all p<0.0001). For true SSN growth, GGNs usually take a median follow-up of 7 years to grow; PSNs usually take a median follow-up of 3 years to grow. For substantial SSN growth, GGNs usually take a median follow-up of 9 years to grow; PSNs usually take a median follow-up of 3 years to grow. For stage shift, GGNs usually take a median follow-up of 12 years to grow; PSNs usually take a median follow-up of 9 years to grow. CONCLUSIONS: The natural course in terms of true growth, substantial growth, and stage shift differed significantly according to their nodule type, which could contribute to the development of follow-up guidelines and management strategy of pulmonary SSNs.

Ethics of Artificial Intelligence in Radiology: Summary of the Joint European and North American Multisociety Statement.

This is a condensed summary of an international multisociety statement on ethics of artificial intelligence (AI) in radiology produced by the ACR, European Society of Radiology, RSNA, Society for Imaging Informatics in Medicine, European Society of Medical Imaging Informatics, Canadian Association of Radiologists, and American Association of Physicists in Medicine. AI has great potential to increase efficiency and accuracy throughout radiology, but it also carries inherent pitfalls and biases. Widespread use of AI-based intelligent and autonomous systems in radiology can increase the risk of systemic errors with high consequence and highlights complex ethical and societal issues. Currently, there is little experience using AI for patient care in diverse clinical settings. Extensive research is needed to understand how to best deploy AI in clinical practice. This statement highlights our consensus that ethical use of AI in radiology should promote well-being, minimize harm, and ensure that the benefits and harms are distributed among stakeholders in a just manner. We believe AI should respect human rights and freedoms, including dignity and privacy. It should be designed for maximum transparency and dependability. Ultimate responsibility and accountability for AI remains with its human designers and operators for the foreseeable future. The radiology community should start now to develop codes of ethics and practice for AI that promote any use that helps patients and the common good and should block use of radiology data and algorithms for financial gain without those two attributes.