Localization of contrast-enhanced breast lesions in ultrafast screening MRI using deep convolutional neural networks.

OBJECTIVES: To develop a deep learning-based method for contrast-enhanced breast lesion detection in ultrafast screening MRI. MATERIALS AND METHODS: A total of 837 breast MRI exams of 488 consecutive patients were included. Lesion's location was independently annotated in the maximum intensity projection (MIP) image of the last time-resolved angiography with stochastic trajectories (TWIST) sequence for each individual breast, resulting in 265 lesions (190 benign, 75 malignant) in 163 breasts (133 women). YOLOv5 models were fine-tuned using training sets containing the same number of MIP images with and without lesions. A long short-term memory (LSTM) network was employed to help reduce false positive predictions. The integrated system was then evaluated on test sets containing enriched uninvolved breasts during cross-validation to mimic the performance in a screening scenario. RESULTS: In five-fold cross-validation, the YOLOv5x model showed a sensitivity of 0.95, 0.97, 0.98, and 0.99, with 0.125, 0.25, 0.5, and 1 false positive per breast, respectively. The LSTM network reduced 15.5% of the false positive prediction from the YOLO model, and the positive predictive value was increased from 0.22 to 0.25. CONCLUSIONS: A fine-tuned YOLOv5x model can detect breast lesions on ultrafast MRI with high sensitivity in a screening population, and the output of the model could be further refined by an LSTM network to reduce the amount of false positive predictions. CLINICAL RELEVANCE STATEMENT: The proposed integrated system would make the ultrafast MRI screening process more effective by assisting radiologists in prioritizing suspicious examinations and supporting the diagnostic workup. KEY POINTS: • Deep convolutional neural networks could be utilized to automatically pinpoint breast lesions in screening MRI with high sensitivity. • False positive predictions significantly increased when the detection models were tested on highly unbalanced test sets with more normal scans. • Dynamic enhancement patterns of breast lesions during contrast inflow learned by the long short-term memory networks helped to reduce false positive predictions.

Reduced Lung-Cancer Mortality with Volume CT Screening in a Randomized Trial.

BACKGROUND: There are limited data from randomized trials regarding whether volume-based, low-dose computed tomographic (CT) screening can reduce lung-cancer mortality among male former and current smokers. METHODS: A total of 13,195 men (primary analysis) and 2594 women (subgroup analyses) between the ages of 50 and 74 were randomly assigned to undergo CT screening at T0 (baseline), year 1, year 3, and year 5.5 or no screening. We obtained data on cancer diagnosis and the date and cause of death through linkages with national registries in the Netherlands and Belgium, and a review committee confirmed lung cancer as the cause of death when possible. A minimum follow-up of 10 years until December 31, 2015, was completed for all participants. RESULTS: Among men, the average adherence to CT screening was 90.0%. On average, 9.2% of the screened participants underwent at least one additional CT scan (initially indeterminate). The overall referral rate for suspicious nodules was 2.1%. At 10 years of follow-up, the incidence of lung cancer was 5.58 cases per 1000 person-years in the screening group and 4.91 cases per 1000 person-years in the control group; lung-cancer mortality was 2.50 deaths per 1000 person-years and 3.30 deaths per 1000 person-years, respectively. The cumulative rate ratio for death from lung cancer at 10 years was 0.76 (95% confidence interval [CI], 0.61 to 0.94; P = 0.01) in the screening group as compared with the control group, similar to the values at years 8 and 9. Among women, the rate ratio was 0.67 (95% CI, 0.38 to 1.14) at 10 years of follow-up, with values of 0.41 to 0.52 in years 7 through 9. CONCLUSIONS: In this trial involving high-risk persons, lung-cancer mortality was significantly lower among those who underwent volume CT screening than among those who underwent no screening. There were low rates of follow-up procedures for results suggestive of lung cancer. (Funded by the Netherlands Organization of Health Research and Development and others; NELSON Netherlands Trial Register number, NL580.).

Automated Breast Density Assessment in MRI Using Deep Learning and Radiomics: Strategies for Reducing Inter-Observer Variability.

BACKGROUND: Accurate breast density evaluation allows for more precise risk estimation but suffers from high inter-observer variability. PURPOSE: To evaluate the feasibility of reducing inter-observer variability of breast density assessment through artificial intelligence (AI) assisted interpretation. STUDY TYPE: Retrospective. POPULATION: Six hundred and twenty-one patients without breast prosthesis or reconstructions were randomly divided into training (N = 377), validation (N = 98), and independent test (N = 146) datasets. FIELD STRENGTH/SEQUENCE: 1.5 T and 3.0 T; T1-weighted spectral attenuated inversion recovery. ASSESSMENT: Five radiologists independently assessed each scan in the independent test set to establish the inter-observer variability baseline and to reach a reference standard. Deep learning and three radiomics models were developed for three classification tasks: (i) four Breast Imaging-Reporting and Data System (BI-RADS) breast composition categories (A-D), (ii) dense (categories C, D) vs. non-dense (categories A, B), and (iii) extremely dense (category D) vs. moderately dense (categories A-C). The models were tested against the reference standard on the independent test set. AI-assisted interpretation was performed by majority voting between the models and each radiologist's assessment. STATISTICAL TESTS: Inter-observer variability was assessed using linear-weighted kappa (κ) statistics. Kappa statistics, accuracy, and area under the receiver operating characteristic curve (AUC) were used to assess models against reference standard. RESULTS: In the independent test set, five readers showed an overall substantial agreement on tasks (i) and (ii), but moderate agreement for task (iii). The best-performing model showed substantial agreement with reference standard for tasks (i) and (ii), but moderate agreement for task (iii). With the assistance of the AI models, almost perfect inter-observer variability was obtained for tasks (i) (mean κ = 0.86), (ii) (mean κ = 0.94), and (iii) (mean κ = 0.94). DATA CONCLUSION: Deep learning and radiomics models have the potential to help reduce inter-observer variability of breast density assessment. LEVEL OF EVIDENCE: 3 TECHNICAL EFFICACY: Stage 1.

Using deep learning to safely exclude lesions with only ultrafast breast MRI to shorten acquisition and reading time.

OBJECTIVES: To investigate the feasibility of automatically identifying normal scans in ultrafast breast MRI with artificial intelligence (AI) to increase efficiency and reduce workload. METHODS: In this retrospective analysis, 837 breast MRI examinations performed on 438 women from April 2016 to October 2019 were included. The left and right breasts in each examination were labelled normal (without suspicious lesions) or abnormal (with suspicious lesions) based on final interpretation. Maximum intensity projection (MIP) images of each breast were then used to train a deep learning model. A high sensitivity threshold was calculated based on the detection trade - off (DET) curve on the validation set. The performance of the model was evaluated by receiver operating characteristic analysis of the independent test set. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) with the high sensitivity threshold were calculated. RESULTS: The independent test set consisted of 178 examinations of 149 patients (mean age, 44 years ± 14 [standard deviation]). The trained model achieved an AUC of 0.81 (95% CI: 0.75-0.88) on the independent test set. Applying a threshold of 0.25 yielded a sensitivity of 98% (95% CI: 90%; 100%), an NPV of 98% (95% CI: 89%; 100%), a workload reduction of 15.7%, and a scan time reduction of 16.6%. CONCLUSION: This deep learning model has a high potential to help identify normal scans in ultrafast breast MRI and thereby reduce radiologists' workload and scan time. KEY POINTS: • Deep learning in TWIST may eliminate the necessity of additional sequences for identifying normal breasts during MRI screening. • Workload and scanning time reductions of 15.7% and 16.6%, respectively, could be achieved with the cost of 1 (1 of 55) false negative prediction.

Facilitating standardized COVID-19 suspicion prediction based on computed tomography radiomics in a multi-demographic setting.

OBJECTIVE: To develop an automatic COVID-19 Reporting and Data System (CO-RADS)-based classification in a multi-demographic setting. METHODS: This multi-institutional review boards-approved retrospective study included 2720 chest CT scans (mean age, 58 years [range 18-100 years]) from Italian and Russian patients. Three board-certified radiologists from three countries assessed randomly selected subcohorts from each population and provided CO-RADS-based annotations. CT radiomic features were extracted from the selected subcohorts after preprocessing steps like lung lobe segmentation and automatic noise reduction. We compared three machine learning models, logistic regression (LR), multilayer perceptron (MLP), and random forest (RF) for the automated CO-RADS classification. Model evaluation was carried out in two scenarios, first, training on a mixed multi-demographic subcohort and testing on an independent hold-out dataset. In the second scenario, training was done on a single demography and externally validated on the other demography. RESULTS: The overall inter-observer agreement for the CO-RADS scoring between the radiologists was substantial (k = 0.80). Irrespective of the type of validation test scenario, suspected COVID-19 CT scans were identified with an accuracy of 84%. SHapley Additive exPlanations (SHAP) interpretation showed that the "wavelet_(LH)_GLCM_Imc1" feature had a positive impact on COVID prediction both with and without noise reduction. The application of noise reduction improved the overall performance between the classifiers for all types. CONCLUSION: Using an automated model based on the COVID-19 Reporting and Data System (CO-RADS), we achieved clinically acceptable performance in a multi-demographic setting. This approach can serve as a standardized tool for automated COVID-19 assessment. KEYPOINTS: • Automatic CO-RADS scoring of large-scale multi-demographic chest CTs with mean AUC of 0.93 ± 0.04. • Validation procedure resembles TRIPOD 2b and 3 categories, enhancing the quality of experimental design to test the cross-dataset domain shift between institutions aiding clinical integration. • Identification of COVID-19 pneumonia in the presence of community-acquired pneumonia and other comorbidities with an AUC of 0.92.

2D Gait Skeleton Data Normalization for Quantitative Assessment of Movement Disorders from Freehand Single Camera Video Recordings.

Overlapping phenotypic features between Early Onset Ataxia (EOA) and Developmental Coordination Disorder (DCD) can complicate the clinical distinction of these disorders. Clinical rating scales are a common way to quantify movement disorders but in children these scales also rely on the observer's assessment and interpretation. Despite the introduction of inertial measurement units for objective and more precise evaluation, special hardware is still required, restricting their widespread application. Gait video recordings of movement disorder patients are frequently captured in routine clinical settings, but there is presently no suitable quantitative analysis method for these recordings. Owing to advancements in computer vision technology, deep learning pose estimation techniques may soon be ready for convenient and low-cost clinical usage. This study presents a framework based on 2D video recording in the coronal plane and pose estimation for the quantitative assessment of gait in movement disorders. To allow the calculation of distance-based features, seven different methods to normalize 2D skeleton keypoint data derived from pose estimation using deep neural networks applied to freehand video recording of gait were evaluated. In our experiments, 15 children (five EOA, five DCD and five healthy controls) were asked to walk naturally while being videotaped by a single camera in 1280 × 720 resolution at 25 frames per second. The high likelihood of the prediction of keypoint locations (mean = 0.889, standard deviation = 0.02) demonstrates the potential for distance-based features derived from routine video recordings to assist in the clinical evaluation of movement in EOA and DCD. By comparison of mean absolute angle error and mean variance of distance, the normalization methods using the Euclidean (2D) distance of left shoulder and right hip, or the average distance from left shoulder to right hip and from right shoulder to left hip were found to better perform for deriving distance-based features and further quantitative assessment of movement disorders.

Focal pericoronary adipose tissue attenuation is related to plaque presence, plaque type, and stenosis severity in coronary CTA.

OBJECTIVES: To investigate the association of pericoronary adipose tissue mean attenuation (PCATMA) with coronary artery disease (CAD) characteristics on coronary computed tomography angiography (CCTA). METHODS: We retrospectively investigated 165 symptomatic patients who underwent third-generation dual-source CCTA at 70kVp: 93 with and 72 without CAD (204 arteries with plaque, 291 without plaque). CCTA was evaluated for presence and characteristics of CAD per artery. PCATMA was measured proximally and across the most severe stenosis. Patient-level, proximal PCATMA was defined as the mean of the proximal PCATMA of the three main coronary arteries. Analyses were performed on patient and vessel level. RESULTS: Mean proximal PCATMA was -96.2 ± 7.1 HU and -95.6 ± 7.8HU for patients with and without CAD (p = 0.644). In arteries with plaque, proximal and lesion-specific PCATMA was similar (-96.1 ± 9.6 HU, -95.9 ± 11.2 HU, p = 0.608). Lesion-specific PCATMA of arteries with plaque (-94.7 HU) differed from proximal PCATMA of arteries without plaque (-97.2 HU, p = 0.015). Minimal stenosis showed higher lesion-specific PCATMA (-94.0 HU) than severe stenosis (-98.5 HU, p = 0.030). Lesion-specific PCATMA of non-calcified, mixed, and calcified plaque was -96.5 HU, -94.6 HU, and -89.9 HU (p = 0.004). Vessel-based total plaque, lipid-rich necrotic core, and calcified plaque burden showed a very weak to moderate correlation with proximal PCATMA. CONCLUSIONS: Lesion-specific PCATMA was higher in arteries with plaque than proximal PCATMA in arteries without plaque. Lesion-specific PCATMA was higher in non-calcified and mixed plaques compared to calcified plaques, and in minimal stenosis compared to severe; proximal PCATMA did not show these relationships. This suggests that lesion-specific PCATMA is related to plaque development and vulnerability. KEY POINTS: • In symptomatic patients undergoing CCTA at 70 kVp, PCATMA was higher in coronary arteries with plaque than those without plaque. • PCATMA was higher for non-calcified and mixed plaques compared to calcified plaques, and for minimal stenosis compared to severe stenosis. • In contrast to PCATMA measurement of the proximal vessels, lesion-specific PCATMA showed clear relationships with plaque presence and stenosis degree.

Clinical and Radiologic Predictors of Parastomal Hernia Development After End Colostomy.

OBJECTIVE. Parastomal hernia (PSH) is a common complication that can occur after end colostomy and may result in considerable morbidity. To select the best candidates for prophylactic measures, knowledge of preoperative PSH predictors is important. This study aimed to determine the value of clinical parameters, preoperative CT-based body metrics, and size of the abdominal wall defect created during end colostomy and measured at postoperative CT for predicting PSH development. MATERIALS AND METHODS. Sixty-five patients who underwent permanent end colostomy with at least 1 year of follow-up were included. On preoperative CT, waist circumference, abdominal wall and psoas muscle indexes, rectus abdominis muscle diameter and diastasis, intra- and extraabdominal fat mass, and presence of other hernias were assessed. On postoperative CT, size of the abdominal wall defect and the presence of PSH were determined. To identify independent predictors of PSH development, univariate analysis with the Kaplan-Meier method and multivariate Cox regression analysis were performed. RESULTS. PSH developed after surgery in 30 patients (46%). Three independent risk factors were identified: chronic obstructive pulmonary disease (COPD) as a comorbidity (hazard ratio [HR], 6.4; 95% CI, 1.9-22.0; p = 0.003), operation time longer than 395 minutes (HR, 3.9; 95% CI, 1.5-10.0; p = 0.005), and maximum aperture diameter of more than 34 mm (HR, 5.2; 95% CI, 2.1-12.7; p < 0.001). PSH developed in all nine patients with a maximum abdominal wall defect diameter of more than 50 mm at the ostomy site. CONCLUSION. COPD, longer operation time, and larger abdominal wall defect at the colostomy site can predict PSH development. Intraoperative creation of an abdominal wall ostomy opening that is more than 34 mm in diameter should be avoided.

Towards reference values of pericoronary adipose tissue attenuation: impact of coronary artery and tube voltage in coronary computed tomography angiography.

OBJECTIVES: To determine normal pericoronary adipose tissue mean attenuation (PCATMA) values for left the anterior descending (LAD), left circumflex (LCX), and right coronary artery (RCA) in patients without plaques on coronary CT angiography (cCTA), taking into account tube voltage influence. METHODS: This retrospective study included 192 patients (76 (39.6%) men; median age 49 years (range, 19-79)) who underwent cCTA with third-generation dual-source CT for the suspicion of CAD between 2015 and 2017. We selected patients without plaque on cCTA. PCATMA was measured semi-automatically on cCTA images in the proximal segment of the three main coronary arteries with 10 mm length. Paired t-testing was used to compare PCATMA between combinations of two coronary arteries within each patient, and one-way ANOVA testing was used to compare PCATMA in different kV groups. RESULTS: The overall mean ± standard deviation (SD) PCATMA was - 90.3 ± 11.1 HU. PCATMA in men was higher than that in women: - 88.5 ± 10.5 HU versus - 91.5 ± 11.3 HU (p = 0.001). PCATMA of LAD, LCX, and RCA was - 92.4 ± 11.6 HU, - 88.4 ± 9.9 HU, and - 90.2 ± 11.4 HU, respectively. Pairwise comparison of the arteries showed significant difference in PCATMA: LAD and LCX (p < 0.001), LAD and RCA (p = 0.009), LCX and RCA (p = 0.033). PCATMA of the 70 kV, 80 kV, 90 kV, 100 kV, and 120 kV groups was - 95.6 ± 9.6 HU, - 90.2 ± 11.5 HU, - 87.3 ± 9.9 HU, - 82.7 ± 6.2 HU, and - 79.3 ± 6.8 HU, respectively (p < 0.001). CONCLUSIONS: In patients without plaque on cCTA, PCATMA varied by tube voltage, with minor differences in PCATMA between coronary arteries (LAD, LCX, RCA). PCATMA values need to be interpreted taking into account tube voltage setting. KEY POINTS: • In patients without plaque on cCTA, PCATMA differs slightly by coronary artery (LAD, LCX, RCA). • Tube voltage of cCTA affects PCATMA measurement, with mean PCATMA increasing linearly with increasing kV. • For longitudinal cCTA analysis of PCATMA , the use of equal kV setting is strongly recommended.

Early imaging biomarkers of lung cancer, COPD and coronary artery disease in the general population: rationale and design of the ImaLife (Imaging in Lifelines) Study.

Lung cancer, chronic obstructive pulmonary disease (COPD), and coronary artery disease (CAD) are expected to cause most deaths by 2050. State-of-the-art computed tomography (CT) allows early detection of lung cancer and simultaneous evaluation of imaging biomarkers for the early stages of COPD, based on pulmonary density and bronchial wall thickness, and of CAD, based on the coronary artery calcium score (CACS), at low radiation dose. To determine cut-off values for positive tests for elevated risk and presence of disease is one of the major tasks before considering implementation of CT screening in a general population. The ImaLife (Imaging in Lifelines) study, embedded in the Lifelines study, is designed to establish the reference values of the imaging biomarkers for the big three diseases in a well-defined general population aged 45 years and older. In total, 12,000 participants will undergo CACS and chest acquisitions with latest CT technology. The estimated percentage of individuals with lung nodules needing further workup is around 1-2%. Given the around 10% prevalence of COPD and CAD in the general population, the expected number of COPD and CAD is around 1000 each. So far, nearly 4000 participants have been included. The ImaLife study will allow differentiation between normal aging of the pulmonary and cardiovascular system and early stages of the big three diseases based on low-dose CT imaging. This information can be finally integrated into personalized precision health strategies in the general population.

Promises of artificial intelligence in neuroradiology: a systematic technographic review.

PURPOSE: To conduct a systematic review of the possibilities of artificial intelligence (AI) in neuroradiology by performing an objective, systematic assessment of available applications. To analyse the potential impacts of AI applications on the work of neuroradiologists. METHODS: We identified AI applications offered on the market during the period 2017-2019. We systematically collected and structured information in a relational database and coded for the characteristics of the applications, their functionalities for the radiology workflow and their potential impacts in terms of 'supporting', 'extending' and 'replacing' radiology tasks. RESULTS: We identified 37 AI applications in the domain of neuroradiology from 27 vendors, together offering 111 functionalities. The majority of functionalities 'support' radiologists, especially for the detection and interpretation of image findings. The second-largest group of functionalities 'extends' the possibilities of radiologists by providing quantitative information about pathological findings. A small but noticeable portion of functionalities seek to 'replace' certain radiology tasks. CONCLUSION: Artificial intelligence in neuroradiology is not only in the stage of development and testing but also available for clinical practice. The majority of functionalities support radiologists or extend their tasks. None of the applications can replace the entire radiology profession, but a few applications can do so for a limited set of tasks. Scientific validation of the AI products is more limited than the regulatory approval.

Design, Implementation, and Validation of a Pulsatile Heart Phantom Pump.

The developments in Computed Tomography (CT) and Magnetic Resonance allow visualization of blood flow in vivo using these techniques. However, validation tests are needed to determine a gold standard. For the validation tests, controllable systems that can generate pulsatile flow are needed. In this study, we aimed to develop an affordable pulsatile pump and an artificial circulatory system to simulate the blood flow for validation purposes. Initially, the prerequisites for the phantom were pulsating flow output equal to that of the human cardiac pulse pattern; the flow pattern of the mimicked cardiac output should be equal to that of a human, a variable stroke volume (40-120 ml/beat), and a variable heart rate (60-170 bpm). The developed phantom setup was tested with CT scanner. A washout profile was created based on the image intensity of the selected slice. The test was successful for a heart rate of 70 bpm and a stroke volume of 68 ml, but the system failed to work at various heartbeats and stroke volumes. This was due to the problems with software of the microcontroller. As conclusion in this study, we present a proof of concept for a pulsatile heart phantom pump that can be used in validation tests.

Assessment of Dynamic Change of Coronary Artery Geometry and Its Relationship to Coronary Artery Disease, Based on Coronary CT Angiography.

To investigate the relationship between dynamic changes of coronary artery geometry and coronary artery disease (CAD) using computed tomography (CT). Seventy-one patients underwent coronary CT angiography with retrospective electrocardiographic gating. End-systolic (ES) and end-diastolic (ED) phases were automatically determined by dedicated software. Centerlines were extracted for the right and left coronary artery. Differences between ES and ED curvature and tortuosity were determined. Associations of change in geometrical parameters with plaque types and degree of stenosis were investigated using linear mixed models. The differences in number of inflection points were analyzed using Wilcoxon signed-rank tests. Tests were done on artery and segment level. One hundred thirty-seven arteries (64.3%) and 456 (71.4%) segments were included. Curvature was significantly higher in ES than in ED phase for arteries (p = 0.002) and segments (p < 0.001). The difference was significant only at segment level for tortuosity (p = 0.005). Number of inflection points was significantly higher in ES phase on both artery and segment level (p < 0.001). No significant relationships were found between degree of stenosis and plaque types and dynamic change in geometrical parameters. Non-invasive imaging by cardiac CT can quantify change in geometrical parameters of the coronary arteries during the cardiac cycle. Dynamic change of vessel geometry through the cardiac cycle was not found to be related to the presence of CAD.

Contextual Structured Reporting in Radiology: Implementation and Long-Term Evaluation in Improving the Communication of Critical Findings.

Structured reporting contributes to the completeness of radiology reports and improves quality. Both the content and the structure are essential for successful implementation of structured reporting. Contextual structured reporting is tailored to a specific scenario and can contain information retrieved from the context. Critical findings detected by imaging need urgent communication to the referring physician. According to guidelines, the occurrence of this communication should be documented in the radiology reports and should contain when, to whom and how was communicated. In free-text reporting, one or more of these required items might be omitted. We developed a contextual structured reporting template to ensure complete documentation of the communication of critical findings. The WHEN and HOW items were included automatically, and the insertion of the WHO-item was facilitated by the template. A pre- and post-implementation study demonstrated a substantial improvement in guideline adherence. The template usage improved in the long-term post-implementation study compared with the short-term results. The two most often occurring categories of critical findings are "infection / inflammation" and "oncology", corresponding to the a large part of urgency level 2 (to be reported within 6 h) and level 3 (to be reported within 6 days), respectively. We conclude that contextual structured reporting is feasible for required elements in radiology reporting and for automated insertion of context-dependent data. Contextual structured reporting improves guideline adherence for communication of critical findings.

Persisting new nodules in incidence rounds of the NELSON CT lung cancer screening study.

BACKGROUND: The US guidelines recommend low-dose CT (LDCT) lung cancer screening for high-risk individuals. New solid nodules after baseline screening are common and have a high lung cancer probability. Currently, no evidence exists concerning the risk stratification of non-resolving new solid nodules at first LDCT screening after initial detection. METHODS: In the Dutch-Belgian Randomized Lung Cancer Screening (NELSON) trial, 7295 participants underwent the second and 6922 participants the third screening round. We included participants with solid nodules that were registered as new or <15 mm³ (study detection limit) at previous screens and received additional screening after initial detection, thereby excluding high-risk nodules according to the NELSON management protocol (nodules ≥500 mm3). RESULTS: Overall, 680 participants with 1020 low-risk and intermediate-risk new solid nodules were included. A total of 562 (55%) new solid nodules were resolving, leaving 356 (52%) participants with a non-resolving new solid nodule, of whom 25 (7%) were diagnosed with lung cancer. At first screening after initial detection, volume doubling time (VDT), volume, and VDT combined with a predefined ≥200 mm3 volume cut-off had high discrimination for lung cancer (VDT, area under the curve (AUC): 0.913; volume, AUC: 0.875; VDT and ≥200 mm3 combination, AUC: 0.939). Classifying a new solid nodule with either ≤590 days VDT or ≥200 mm3 volume positive provided 100% sensitivity, 84% specificity and 27% positive predictive value for lung cancer. CONCLUSIONS: More than half of new low-risk and intermediate-risk solid nodules in LDCT lung cancer screening resolve. At follow-up, growth assessment potentially combined with a volume limit can be used for risk stratification. TRIAL REGISTRATION NUMBER: ISRCTN63545820; pre-results.

Convolutional neural network to predict the local recurrence of giant cell tumor of bone after curettage based on pre-surgery magnetic resonance images.

OBJECTIVE: To predict the local recurrence of giant cell bone tumors (GCTB) on MR features and the clinical characteristics after curettage using a deep convolutional neural network (CNN). METHODS: MR images were collected from 56 patients with histopathologically confirmed GCTB after curettage who were followed up for 5.8 years (range, 2.0 to 9.5 years). The inception v3 CNN architecture was fine-tuned by two categories of the MR datasets (recurrent and non-recurrent GCTB) obtained through data augmentation and was validated using fourfold cross-validation to evaluate its generalization ability. Twenty-eight cases (50%) were chosen as the training dataset for the CNN and four radiologists, while the remaining 28 cases (50%) were used as the test dataset. A binary logistic regression model was established to predict recurrent GCTB by combining the CNN prediction and patient features (age and tumor location). Accuracy and sensitivity were used to evaluate the prediction performance. RESULTS: When comparing the CNN, CNN regression, and radiologists, the accuracies of the CNN and CNN regression models were 75.5% (95% CI 55.1 to 89.3%) and 78.6% (59.0 to 91.7%), respectively, which were higher than the 64.3% (44.1 to 81.4%) accuracy of the radiologists. The sensitivities were 85.7% (42.1 to 99.6%) and 87.5% (47.3 to 99.7%), respectively, which were higher than the 58.3% (27.7 to 84.8%) sensitivity of the radiologists (p < 0.05). CONCLUSION: The CNN has the potential to predict recurrent GCTB after curettage. A binary regression model combined with patient characteristics improves its prediction accuracy. KEY POINTS: • Convolutional neural network (CNN) can be trained successfully on a limited number of pre-surgery MR images, by fine-tuning a pre-trained CNN architecture. • CNN has an accuracy of 75.5% to predict post-surgery recurrence of giant cell tumors of bone, which surpasses the 64.3% accuracy of human observation. • A binary logistic regression model combining CNN prediction rate, patient age, and tumor location improves the accuracy to predict post-surgery recurrence of giant cell bone tumors to 78.6%.

Disagreement of diameter and volume measurements for pulmonary nodule size estimation in CT lung cancer screening.

We studied 2240 indeterminate solid nodules (volume 50-500mm3) to determine the correlation of diameter and semi-automated volume measurements for pulmonary nodule size estimation. Intra-nodular diameter variation, defined as maximum minus minimum diameter through the nodule's center, varied by 2.8 mm (median, IQR:2.2-3.7 mm), so above the 1.5 mm cutoff for nodule growth used in Lung CT Screening Reporting and Data System (Lung-RADS). Using mean or maximum axial diameter to assess nodule volume led to a substantial mean overestimation of nodule volume of 47.2% and 85.1%, respectively, compared to semi-automated volume. Thus, size of indeterminate nodules is poorly represented by diameter. TRIAL REGISTRATION NUMBER: Pre-results, ISRCTN63545820.

Airway wall thickness on HRCT scans decreases with age and increases with smoking.

BACKGROUND: To investigate if age, gender and smoking are associated with airway wall thickness (AWT) measured by high resolution computed tomography (HRCT) and if higher AWT is associated with lower levels of pulmonary function in healthy current- and never-smokers with a wide age range. METHODS: HRCT scans were performed in 99 subjects (48 never- and 51 current-smokers, median age 39 years [IQR 22 - 54], 57% males). The AWT at an internal perimeter of 10 mm (AWT Pi10) was calculated as an overall measurement of AWT, based on all measurements throughout the lungs. Extensive pulmonary function testing was performed in all subjects. RESULTS: Higher age was associated with a lower AWT Pi10 (b = -0.003, p < 0.001). Current-smokers had a higher AWT Pi10 than never-smokers (mean 0.49 mm versus 0.44 mm, p = 0.022). In multivariate analysis, age and current-smoking were independently associated with AWT Pi10 (age b = -0.002, p < 0.001, current-smoking b = 0.041, p = 0.021), whereas gender was not (b = 0.011, p = 0.552). Higher AWT Pi10 was associated with a lower FEV1, FEV1/FVC, FEF25-75 and higher R5, R20 and X5. CONCLUSIONS: AWT decreases with higher age, possibly reflecting structural changes of the airways. Additionally, current-smokers have a higher AWT, possibly due to remodeling or inflammation. Finally, higher AWT is associated with a lower level of pulmonary function, even in this population of healthy subjects. TRIAL REGISTRATION: This Study was registered at www.clinicaltrials.gov with number NCT00848406 on 19 February 2009.

Occurrence and lung cancer probability of new solid nodules at incidence screening with low-dose CT: analysis of data from the randomised, controlled NELSON trial.

BACKGROUND: US guidelines now recommend lung cancer screening with low-dose CT for high-risk individuals. Reports of new nodules after baseline screening have been scarce and are inconsistent because of differences in definitions used. We aimed to identify the occurrence of new solid nodules and their probability of being lung cancer at incidence screening rounds in the Dutch-Belgian Randomized Lung Cancer Screening Trial (NELSON). METHODS: In the ongoing, multicentre, randomised controlled NELSON trial, between Dec 23, 2003, and July 6, 2006, 15 822 participants who had smoked at least 15 cigarettes a day for more than 25 years or ten cigarettes a day for more than 30 years and were current smokers, or had quit smoking less than 10 years ago, were enrolled and randomly assigned to receive either screening with low-dose CT (n=7915) or no screening (n=7907). From Jan 28, 2004, to Dec 18, 2006, 7557 individuals underwent baseline screening with low-dose CT; 7295 participants underwent second and third screening rounds. We included all participants with solid non-calcified nodules, registered by the NELSON radiologists as new or smaller than 15 mm(3) (study detection limit) at previous screens. Nodule volume was generated semiautomatically by software. We calculated the maximum volume doubling time for nodules with an estimated percentage volume change of 25% or more, representing the minimum growth rate for the time since the previous scan. Lung cancer diagnosis was based on histology, and benignity was based on histology or stable size for at least 2 years. The NELSON trial is registered at trialregister.nl, number ISRCTN63545820. FINDINGS: We analysed data for participants with at least one solid non-calcified nodule at the second or third screening round. In the two incidence screening rounds, the NELSON radiologists registered 1222 new solid nodules in 787 (11%) participants. A new solid nodule was lung cancer in 49 (6%) participants with new solid nodules and, in total, 50 lung cancers were found, representing 4% of all new solid nodules. 34 (68%) lung cancers were diagnosed at stage I. Nodule volume had a high discriminatory power (area under the receiver operating curve 0·795 [95% CI 0·728-0·862]; p<0·0001). Nodules smaller than 27 mm(3) had a low probability of lung cancer (two [0·5%] of 417 nodules; lung cancer probability 0·5% [95% CI 0·0-1·9]), nodules with a volume of 27 mm(3) up to 206 mm(3) had an intermediate probability (17 [3·1%] of 542 nodules; lung cancer probability 3·1% [1·9-5·0]), and nodules of 206 mm(3) or greater had a high probability (29 [16·9%] of 172 nodules; lung cancer probability 16·9% [12·0-23·2]). A volume cutoff value of 27 mm(3) or greater had more than 95% sensitivity for lung cancer. INTERPRETATION: Our study shows that new solid nodules are detected at each screening round in 5-7% of individuals who undergo screening for lung cancer with low-dose CT. These new nodules have a high probability of malignancy even at a small size. These findings should be considered in future screening guidelines, and new solid nodules should be followed up more aggressively than nodules detected at baseline screening. FUNDING: Zorgonderzoek Nederland Medische Wetenschappen and Koningin Wilhelmina Fonds Kankerbestrijding.

Quantitative STIR of muscle for monitoring nerve regeneration.

PURPOSE: To assess whether short tau inversion recovery (STIR) MRI sequences can provide a tool for monitoring peripheral nerve regeneration, by comparing signal intensity changes in reinnervated muscle over time, and to determine potential clinical time points for monitoring. MATERIALS AND METHODS: For this prospective study, 29 patients with complete traumatic transection of the ulnar or median nerves in the forearm were followed up to 45 months postsurgery. Standardized 1.5 Tesla STIR-MRI scans of hand muscles were obtained at fixed time intervals. Muscle signal intensities were measured semi-quantitatively and correlated to functional outcome. RESULTS: For the patients with good function recovery, mean signal intensity ratios of 1.179 ± 0.039, 1.304 ± 0.180, 1.154 ± 0.121, 1.105 ± 0.046 and 1.038 ± 0.047 were found at 1-, 3-, 6-, 9-, and 12-month follow-up, respectively. In the group with poor function recovery, ratios of 1.240 ± 0.069, 1.374 ± 0.144, 1.407 ± 0.127, 1.386 ± 0.128 and 1.316 ± 0.116 were found. Comparing the groups showed significant differences from 6 months onward (P < 0.001), with normalizing signal intensities in the group with good function recovery and sustained elevated signal intensity in the group with poor function recovery. CONCLUSION: MRI of muscle can be used as a tool for monitoring motor nerve regeneration, by comparing STIR muscle signal intensities over time. A decrease in signal intensity ratio of 50% (as compared to the initial increase) seems to predict good function recovery. Long-term follow-up shows that STIR MRI can be used for at least 15 months after nerve transection to differentiate between denervated and (re)innervated muscles. J. Magn. Reson. Imaging 2016;44:401-410.