Detection of acute rib fractures on CT images with convolutional neural networks: effect of location and type of fracture and reader’s experience

Emergency Radiology - The evaluation of all ribs on thin-slice CT images is time consuming and it can be difficult to accurately assess the location and type of rib fracture in an emergency. The...     

Detection of metallic objects on digital radiographs with convolutional neural networks: A MRI screening tool

IntroductionScreening for metallic implants and foreign bodies before magnetic resonance imaging (MRI) examinations, are crucial for patient safety. History of health are supplied by the patient, a family member, screening of electronic health records or the picture and archive systems (PACS). PACS securely store and transmits digital radiographs (DR) and related reports with patient information. Convolutional neural networks (CNN) can be used to detect metallic objects in DRs stored in PACS. This study evaluates the accuracy of CNNs in the detection of metallic objects on DRs as an MRI screening tool.MethodsThe musculoskeletal radiographs (MURA) dataset consisting of 14.863 upper extremity studies were stratified into datasets with and without metal. For each anatomical region: Elbow, finger, hand, humerus, forearm, shoulder and wrist we trained and validated CNN algorithms to classify radiographs with and without metal. Algorithm performance was evaluated with area under the receiver-operating curve (AUC), sensitivity, specificity, predictive values and accuracies compared with a reference standard of manually labelling.ResultsSensitivities, specificities and area under the ROC-curves (AUC) for the six anatomic regions ranged from 85.33% (95% CI: 78.64%–90.57%) to 100.00% (95% CI: 98.16%–100.00%), 75.44% (95% CI: 62.24%–85.87%) to 93.57% (95% CI: 88.78%–96.75%) and 0.95 to 0.99, respectively.ConclusionCNN algorithms classify DRs with metallic objects for six different anatomic regions with near-perfect accuracy. The rapid and iterative capability of the algorithms allows for scalable expansion and as a substitute MRI screening tool for metallic objects.Implications for practiceAll CNNs would be able to assist in metal detection of digital radiographs prior to MRI, an substantially decrease screening time.     

Missed rib fractures on evaluation of initial chest CT for trauma patients: pattern analysis and diagnostic value of coronal multiplanar reconstruction images with multidetector row CT

Objective The objective of this study was to review the prevalence and radiological features of rib fractures missed on initial chest CT evaluation, and to examine the diagnostic value of additional coronal images in a large series of trauma patients. Methods 130 patients who presented to an emergency room for blunt chest trauma underwent multidetector row CT of the thorax within the first hour during their stay, and had follow-up CT or bone scans as diagnostic gold standards. Images were evaluated on two separate occasions: once with axial images and once with both axial and coronal images. The detection rates of missed rib fractures were compared between readings using a non-parametric method of clustered data. In the cases of missed rib fractures, the shapes, locations and associated fractures were evaluated. Results 58 rib fractures were missed with axial images only and 52 were missed with both axial and coronal images (p=0.088). The most common shape of missed rib fractures was buckled (56.9%), and the anterior arc (55.2%) was most commonly involved. 21 (36.2%) missed rib fractures had combined fractures on the same ribs, and 38 (65.5%) were accompanied by fracture on neighbouring ribs. Conclusion Missed rib fractures are not uncommon, and radiologists should be familiar with buckle fractures, which are frequently missed. Additional coronal imagescan be helpful in the diagnosis of rib fractures that are not seen on axial images.     

Cognitive signature of brain FDG PET based on deep learning: domain transfer from Alzheimer’s disease to Parkinson’s disease

European Journal of Nuclear Medicine and Molecular Imaging - Although functional brain imaging has been used for the early and objective assessment of cognitive dysfunction, there is a lack of...     

Incidental vertebral fractures on multidetector CT images of the chest: prevalence and recognition

AIM: To investigate the prevalence, demographics, fracture site, and the rate of vertebral fracture recognition by radiologists on multidetector computed tomography (MDCT) images of the chest. METHOD: The images of 200 consecutive patients who underwent a MDCT examination of the chest over a 4-month period were reviewed. The thoracic spine, viewed using bone windows in sagittal reformats, was reviewed in consensus by three radiologists. Vertebral fractures were assessed using a validated semi-quantitative method. RESULTS: The mean age was 61 years (range 18-92 years); 48% were female. There were 70 (35%) fractures, of which 51 (73%) were mild grade 1 fractures, 13 (19%) moderate grade 2 fractures, and six (9%) severe grade 3 fractures. Fractures commonly affected patients above the age of 50. The most common site was in the lower thoracic region (53%). The overall recognition rate on the radiologists' reports was 6 (9%). CONCLUSIONS: One in three patients who underwent MDCT of the chest had vertebral fractures. There is significant underreporting of these fractures, and the importance of fracture identification should be emphasized to avoid under-diagnosis.     

Appropriateness of a donor liver with respect to macrosteatosis: application of artificial neural networks to US images--initial experience

PURPOSE: To retrospectively compare performance of artificial neural networks (ANNs) applied to ultrasonographic (US) images with that of radiologists for prediction of appropriateness of a donor liver with respect to macrosteatosis before liver transplantation. MATERIALS AND METHODS: Institutional ethics committee approved study; written informed consent was obtained. ANNs, constructed with three-layered 15-neuron back-propagation algorithm, were trained to predict appropriateness of a donor liver with respect to macrosteatosis by using statistically significant laboratory and US parameters derived from univariate analyses, together with correct diagnosis. Input variables for ANNs were alkaline phosphatase, glutamic oxaloacetic transaminase, glutamic pyruvate transaminase, gamma-glutamyltransferase, hepatorenal ratio of echogenicity, and tail area ratio and tail length of portal vein wall echogenicity. Three radiologists graded US images in 94 potential donors (71 men and 23 women) on the basis of four degrees of hepatic steatosis. After training and testing of ANNs, performance of ANNs and radiologists in predicting appropriateness of potential donors was evaluated with receiver operating characteristic (ROC) analysis and compared by means of univariate z score test. RESULTS: Among 94 potential donor livers, 76 were normal or had mild steatosis, and 18 had moderate or severe macrosteatosis at histopathologic examination. Area under ROC curve (Az) of ANNs (Az=0.9673) was significantly greater than that of radiologists (faculty, Az=0.9106, P=.048; fellow, Az= 0.9038, P=.044; resident, Az=0.8931, P=.038). No statistically significant difference in sensitivity for predicting appropriateness as a liver donor with respect to macrosteatosis was found between ANNs (88.9%) and radiologists (P >.05). However, specificity of ANNs (96.1%) was significantly better than that of radiologists (P <.003). CONCLUSION: ANNs might be a useful tool to categorize whether a donor liver is appropriate for transplantation with respect to macrosteatosis on the basis of multiple variables related to laboratory and US features. Further study is needed.     

CT colonography: effect of experience and training on reader performance

The purpose of this paper was to investigate the effect of radiologist experience and increasing exposure to CT colonography on reader performance. Three radiologists of differing general experience (consultant, research fellow, trainee) independently analysed 100 CT colonographic datasets. Readers had no prior experience of CT colonography and received feedback and training after the first 50 cases from an independent experienced radiologist. Diagnostic performance and reporting times were compared for the first and second 50 datasets and compared with the results of a radiologist experienced in CT colonography. Before training only the consultant reader achieved statistical equivalence with the reference standard for detection of larger polyps. After training, detection rates ranged between 25 and 58% for larger polyps. Only the trainee significantly improved after training (P=0.007), with performance of other readers unchanged or even worse. Reporting times following training were reduced significantly for the consultant and fellow (P<0.001 and P=0.03, respectively), but increased for the trainee (P<0.001). In comparison to the consultant reader, the odds of detection of larger polyps was 0.36 (CI 0.16, 0.82) for the fellow and 0.36 (CI 0.14, 0.91) for the trainee. There is considerable variation in the ability to report CT colonography. Prior experience in gastrointestinal radiology is a distinct advantage. Competence cannot be assumed even after directed training via a database of 50 cases.     

Correction to: A phase 2, open-label,multi-center study to evaluate the efficacy and safety of 99m Tc-TRODAT-1 SPECT to detect Parkinson’s disease

The corresponding author of the article would like to remove “Jian Wang” in the author group.     

Detection and characterization of lesions on low-radiation-dose abdominal CT images postprocessed with noise reduction filters

PURPOSE: To assess the effect of noise reduction filters on detection and characterization of lesions on low-radiation-dose abdominal computed tomographic (CT) images. MATERIALS AND METHODS: Low-dose CT images of abdominal lesions in 19 consecutive patients (11 women, eight men; age range, 32-78 years) were obtained at reduced tube currents (120-144 mAs). These baseline low-dose CT images were postprocessed with six noise reduction filters; the resulting postprocessed images were then randomly assorted with baseline images. Three radiologists performed independent evaluation of randomized images for presence, number, margins, attenuation, conspicuity, calcification, and enhancement of lesions, as well as image noise. Side-by-side comparison of baseline images with postprocessed images was performed by using a five-point scale for assessing lesion conspicuity and margins, image noise, beam hardening, and diagnostic acceptability. Quantitative noise and contrast-to-noise ratio were obtained for all liver lesions. Statistical analysis was performed by using the Wilcoxon signed rank test, Student t test, and kappa test of agreement. RESULTS: Significant reduction of noise was observed in images postprocessed with filter F compared with the noise in baseline nonfiltered images (P =.004). Although the number of lesions seen on baseline images and that seen on postprocessed images were identical, lesions were less conspicuous on postprocessed images than on baseline images. A decrease in quantitative image noise and contrast-to-noise ratio for liver lesions was noted with all noise reduction filters. There was good interobserver agreement (kappa = 0.7). CONCLUSION: Although the use of currently available noise reduction filters improves image noise and ameliorates beam-hardening artifacts at low-dose CT, such filters are limited by a compromise in lesion conspicuity and appearance in comparison with lesion conspicuity and appearance on baseline low-dose CT images.     

Reassessment of CT images to improve diagnostic accuracy in patients with suspected acute appendicitis and an equivocal preoperative CT interpretation

Objectives

To identify CT features that discriminate individuals with and without acute appendicitis in patients with equivocal CT findings, and to assess whether knowledge of these findings improves diagnostic accuracy.

Methods

53 patients that underwent appendectomy with an indeterminate preoperative CT interpretation were selected and allocated to an acute appendicitis group or a non-appendicitis group. The 53 CT examinations were reviewed by two radiologists in consensus to identify CT findings that could aid in the discrimination of those with and without appendicitis. In addition, two additional radiologists were then requested to evaluate independently the 53 CT examinations using a 4-point scale, both before and after being informed of the potentially discriminating criteria.

Results

CT findings found to be significantly different in the two groups were; the presence of appendiceal wall enhancement, intraluminal air in appendix, a coexistent inflammatory lesion, and appendiceal wall thickening (P?P?=?0.0193 and P?=?0.0397, respectively).

Conclusions

Knowledge of the identified CT findings was found to improve diagnostic accuracy for acute appendicitis in patients with equivocal CT findings.

Key Points

? Numerous patients with clinically equivocal appendicitis do not have acute appendicitis ? Computed tomography (CT) helps to reduce the negative appendectomy rate ? CT is not always infallible and may also demonstrate indeterminate findings ? However knowledge of significant CT variables can further reduce negative appendectomy rate ? An equivocal CT interpretation of appendicitis should be reassessed with this knowledge     

Sternal fracture in the emergency department: diagnostic value of multidetector CT with sagittal and coronal reconstruction images

Background

Multidetector CT (MDCT) is being increasingly used for patients with traumatic injuries in the emergency room. This is the first study to evaluate the diagnostic performance of MDCT for sternal fracture.

Patients and methods

For 87 patients who had motor vehicle accidents, we evaluated the diagnostic performance of MDCT for the sternal fractures. For 31 patients who underwent both MDCT and lateral radiography for the sternum, we compared the diagnostic performance of two examinations for the evaluation of sternal fracture.

Results

Thirty-two patients had sternal fractures and all the sternal fractures (sensitivity = 100%) were detected on MDCT, especially on the sagittal reconstruction images. However, the axial and coronal images detected 65% and 59% of all sternal fractures, respectively. For 31 patients who underwent both MDCT and lateral radiography for the sternum, MDCT showed superior diagnostic performance compared to that of radiography (accuracy = 97% and 77%, respectively, P = .02). For the one case that showed false positivity for sternal fracture on MDCT due to respiratory artifact, the lateral radiography enabled making the correct diagnosis.

Conclusion

Sternal fracture is frequently seen in patients who have blunt trauma injury secondary to motor vehicle accidents. MDCT, particularly sagittal images detect all of the sternal fractures, is superior to lateral radiography for diagnosis of sternal fracture. In the limited case that CT has severe motion artifact, additional radiography could help the diagnosis of sternal fracture.     

Automatic detection and quantification of ground-glass opacities on high-resolution CT using multiple neural networks: comparison with a density mask

OBJECTIVE: We compared multiple neural networks with a density mask for the automatic detection and quantification of ground-glass opacities on high-resolution CT under clinical conditions. SUBJECTS AND METHODS: Eighty-four patients (54 men and 30 women; age range, 18-82 years; mean age, 49 years) with a total of 99 consecutive high-resolution CT scans were enrolled in the study. The neural network was designed to detect ground-glass opacities with high sensitivity and to omit air-tissue interfaces to increase specificity. The results of the neural network were compared with those of a density mask (thresholds, -750/-300 H), with a radiologist serving as the gold standard. RESULTS: The neural network classified 6% of the total lung area as ground-glass opacities. The density mask failed to detect 1.3%, and this percentage represented the increase in sensitivity that was achieved by the neural network. The density mask identified another 17.3% of the total lung area to be ground-glass opacities that were not detected by the neural network. This area represented the increase in specificity achieved by the neural network. Related to the extent of the ground-glass opacities as classified by the radiologist, the neural network (density mask) reached a sensitivity of 99% (89%), specificity of 83% (55%), positive predictive value of 78% (18%), negative predictive value of 99% (98%), and accuracy of 89% (58%). CONCLUSION: Automatic segmentation and quantification of ground-glass opacities on high-resolution CT by a neural network are sufficiently accurate to be implemented for the preinterpretation of images in a clinical environment; it is superior to a double-threshold density mask.     

Genetic testing in sudden unexpected natural death in the young: New York City Office of Chief Medical Examiner’s experience and perspective

Postmortem genetic testing is a diagnostic tool that is becoming increasingly utilized. The benefits and limitations of genetic testing in cases of sudden, unexpected death in the young (≤ 40 years old) are reviewed from the perspective of the Office of Chief Medical Examiner of the City of New York, whose Molecular Genetics Laboratory, accredited by College of American Pathologists, has had 15 years of postmortem testing experience. Challenges to the interpretation and communication of testing results are highlighted, and opportunities for improving testing yield are discussed for age groups across the lifespan, from infancy to adulthood.