Computerized Diagnostic Assistant for the Automatic Detection of Pneumothorax on Ultrasound: A Pilot Study

Quantitative assessment of pneumothorax by using Shannon entropy of lung ultrasound M-mode image and diaphragmatic excursion based on automated measurement

Background

Lung ultrasound (LUS) and diaphragm ultrasound (DUS) are the appropriate modalities for conservative observation to those patients who are with stable pneumothorax, as well as for the timely detection of life-threatening pneumothorax at any location, due to they are portable, real-time, relatively cost effective, and most important, without radiation exposure. The absence of lung sliding on LUS M-mode images and the abnormality of diaphragmatic excursion (DE) on DUS M-mode images are the most common and novel diagnostic criteria for pneumothorax, respectively. However, visual inspection of M-mode images remains subjective and quantitative analysis of LUS and DUS M-mode images are required.

Methods

Shannon entropy of LUS M-mode image (ShanEnLM) and DE based on the automated measurement (DEAM) are adapted to the objective pneumothorax diagnoses and the severity quantifications in this study. Mild, moderate, and severe pneumothoraces were induced in 24 male New Zealand rabbits through insufflation of room air (5, 10 and 15, and 25 and 40 mL/kg, respectively) into their pleural cavities. In vivo intercostal LUS and subcostal DUS M-mode images were acquired using a point-of-care system for estimating ShanEnLM and DEAM.

Results

ShanEnLM and DEAM as functions of air insufflation volumes exhibited U-shaped curves and were exponentially decreasing, respectively. Either ShanEnLM or DEAM had areas under the receiver operating characteristic curves [95% confidence interval (CI)] of 1.0000 (95% CI: 1.0000-1.0000), 0.9833 (95% CI: 0.9214-1.0000), and 0.9407 (95% CI: 0.8511-1.0000) for differentiating between normal and mild pneumothorax, mild and moderate pneumothoraces, and moderate and severe pneumothoraces, respectively.

Conclusions

Our findings imply that the combination of ShanEnLM and DEAM give the promising potential for pneumothorax quantitative diagnosis.

Classification of lung pathologies in neonates using dual-tree complex wavelet transform

INTRODUCTION

Undiagnosed and untreated lung pathologies are among the leading causes of neonatal deaths in developing countries. Lung Ultrasound (LUS) has been widely accepted as a diagnostic tool for neonatal lung pathologies due to its affordability, portability, and safety. However, healthcare institutions in developing countries lack well-trained clinicians to interpret LUS images, which limits the use of LUS, especially in remote areas. An automated point-of-care tool that could screen and capture LUS morphologies associated with neonatal lung pathologies could aid in rapid and accurate diagnosis.

METHODS

We propose a framework for classifying the six most common neonatal lung pathologies using spatially localized line and texture patterns extracted via 2D dual-tree complex wavelet transform (DTCWT). We acquired 1550 LUS images from 42 neonates with varying numbers of lung pathologies. Furthermore, we balanced our data set to avoid bias towards a pathology class.

RESULTS

Using DTCWT and clinical features as inputs to a linear discriminant analysis (LDA), our approach achieved a per-image cross-validated classification accuracy of 74.39% for the imbalanced data set. Our classification accuracy improved to 92.78% after balancing our data set. Moreover, our proposed framework achieved a maximum per-subject cross-validated classification accuracy of 64.97% with an imbalanced data set while using a balanced data set improves its classification accuracy up to 81.53%.

CONCLUSION

Our work could aid in automating the diagnosis of lung pathologies among neonates using LUS. Rapid and accurate diagnosis of lung pathologies could help to decrease neonatal deaths in healthcare institutions that lack well-trained clinicians, especially in developing countries.

Comparison of Commonly Carried Liquids Against Commercial Ultrasound Gel for Use in the Backcountry Setting

INTRODUCTION

Point-of-care ultrasound (POCUS) is utilized in austere environments because it is lightweight, durable, battery powered, and portable. In austere settings, weight and space constraints are limitations to carrying dedicated ultrasound gel. Few studies have assessed commonly carried liquids as gel alternatives. The study objective was to assess the suitability of common food and personal care products as ultrasound coupling agents compared with that of commercial gel.

METHODS

A noninferiority study compared 9 products to commercial gel. Each substance was independently tested on 2 subjects by 2 sonographers covering 8 standardized ultrasound windows. Clips were recorded, blinded, and independently graded by 2 ultrasound fellowship-trained physicians on the ability to make clinical decisions and technical details, including contrast, resolution, and artifact. A 20% noninferiority margin was set, which correlates to levels considered to be of reliably sufficient quality by American College of Emergency Physicians' guidelines. The substances included water, soap, shampoo, olive oil, energy gel, maple syrup, hand sanitizer, sunscreen, and lotion.

RESULTS

A total of 300 of 318 (94%) clips met the primary endpoint of adequacy to make a clinical decision. All media, except sunscreen, were noninferior to commercial gel in the ability to make a clinical decision (α=0.05). In terms of secondary outcomes, resolution, artifact, and contrast, all substances were noninferior to commercial gel (α=0.05). The sonographers concluded that all gel alternatives' usability performed similarly to commercial gel, with the exception of energy gel.

CONCLUSIONS

Of the 9 substances tested, 8 were noninferior to commercial gels for clinical decisions. Our study indicates that several POCUS gel substitutes are serviceable to produce clinically adequate images.

Detection of pneumothorax on ultrasound using artificial intelligence

BACKGROUND

Ultrasound (US) for the detection of pneumothorax shows excellent sensitivity in the hands of skilled providers. Artificial intelligence may facilitate the movement of US for pneumothorax into the prehospital setting. The large amount of training data required for conventional neural network methodologies has limited their use in US so far.

METHODS

A limited training database was supplied by Defense Advanced Research Projects Agency of 30 patients, 15 cases with pneumothorax and 15 cases without. There were two US videos per patient, of which we were allowed to choose one to train on, so that a limited set of 30 videos were used. Images were annotated for ribs and pleural interface. The software performed anatomic reconstruction to identify the region of interest bounding the pleura. Three neural networks were created to analyze images on a pixel-by-pixel fashion with direct voting determining the outcome. Independent verification and validation was performed on a data set gathered by the Department of Defense.

RESULTS

Anatomic reconstruction with the identification of ribs and pleura was able to be accomplished on all images. On independent verification and validation against the Department of Defense testing data, our program concurred with the SME 80% of the time and achieved a 86% sensitivity (18/21) for pneumothorax and a 75% specificity for the absence of pneumothorax (18/24). Some of the mistakes by our artificial intelligence can be explained by chest wall motion, hepatization of the underlying lung, or being equivocal cases.

CONCLUSION

Using learning with limited labeling techniques, pneumothorax was identified on US with an accuracy of 80%. Several potential improvements are controlling for chest wall motion and the use of longer videos.

LEVEL OF EVIDENCE

Diagnostic Tests; Level III.

Using Ultrasound Image Augmentation and Ensemble Predictions to Prevent Machine-Learning Model Overfitting

Deep learning predictive models have the potential to simplify and automate medical imaging diagnostics by lowering the skill threshold for image interpretation. However, this requires predictive models that are generalized to handle subject variability as seen clinically. Here, we highlight methods to improve test accuracy of an image classifier model for shrapnel identification using tissue phantom image sets. Using a previously developed image classifier neural network-termed ShrapML-blind test accuracy was less than 70% and was variable depending on the training/test data setup, as determined by a leave one subject out (LOSO) holdout methodology. Introduction of affine transformations for image augmentation or MixUp methodologies to generate additional training sets improved model performance and overall accuracy improved to 75%. Further improvements were made by aggregating predictions across five LOSO holdouts. This was done by bagging confidences or predictions from all LOSOs or the top-3 LOSO confidence models for each image prediction. Top-3 LOSO confidence bagging performed best, with test accuracy improved to greater than 85% accuracy for two different blind tissue phantoms. This was confirmed by gradient-weighted class activation mapping to highlight that the image classifier was tracking shrapnel in the image sets. Overall, data augmentation and ensemble prediction approaches were suitable for creating more generalized predictive models for ultrasound image analysis, a critical step for real-time diagnostic deployment.

Prospects of Structural Similarity Index for Medical Image Analysis

An image quality matrix provides a significant principle for objectively observing an image based on an alteration between the original and distorted images. During the past two decades, a novel universal image quality assessment has been developed with the ability of adaptation with human visual perception for measuring the difference of a degraded image from the reference image, namely a structural similarity index. Structural similarity has since been widely used in various sectors, including medical image evaluation. Although numerous studies have reported the use of structural similarity as an evaluation strategy for computer-based medical images, reviews on the prospects of using structural similarity for medical imaging applications have been rare. This paper presents previous studies implementing structural similarity in analyzing medical images from various imaging modalities. In addition, this review describes structural similarity from the perspective of a family’s historical background, as well as progress made from the original to the recent structural similarity, and its strengths and drawbacks. Additionally, potential research directions in applying such similarities related to medical image analyses are described. This review will be beneficial in guiding researchers toward the discovery of potential medical image examination methods that can be improved through structural similarity index.

The Role of Ultrasonography in the Diagnosis and Decision Algorithm for the Management of Pneumothorax after Transbronchial Lung Cryobiopsy

BACKGROUND

Pneumothorax is one of the main complications of transbronchial lung cryobiopsy (TBLC). Chest ultrasound (CUS) is a radiation-free alternative method for pneumothorax detection.

OBJECTIVE

We tested CUS diagnostic accuracy for pneumothorax and assessed its role in the decision algorithm for pneumothorax management. Secondary objectives were to evaluate the post-procedure pneumothorax occurrence and risk factors.

METHODS

Eligible patients underwent TBLC, followed by chest X-ray (CXR) evaluation 2 h after the procedure, as our standard protocol. Bedside CUS was performed within 30 min and 2 h after TBLC. Pneumothorax by CUS was defined by the absence of lung sliding and comet-tail artefacts and confirmed with the stratosphere sign on M-mode. Pneumothorax size was determined through lung point projection on CUS and interpleural distance on CXR and properly managed according to clinical status.

RESULTS

Sixty-seven patients were included. Nineteen pneumothoraces were detected at 2 h after the procedure, of which 8 (42.1%) were already present at the first CUS evaluation. All CXR-detected pneumothoraces had a positive CUS detection. There were 3 discordant cases (κ = 0.88, 95% CI: 0.76-1.00, p < 0.001), which were detected by CUS but not by inspiration CXR. We calculated a specificity of 97.5% (95% CI: 86.8-99.9) and a sensitivity of 100% (95% CI: 87.2-100) for CUS. Pneumothorax rate was higher when biopsies were taken in 2 lobes and if histology had pleural representation. Final diagnosis was achieved in 79.1% of patients, with the most frequent diagnosis being hypersensitivity pneumonitis. Regarding patients with large-volume pneumothorax needing drainage, the rate of detection was similar between CUS and CRX.

CONCLUSION

CUS can replace CXR in detecting the presence of pneumothorax after TBLC, and the lung point site can reliably indicate its size. This useful method optimizes time spent at the bronchology unit and allows immediate response in symptomatic patients, helping to choose optimal treatment strategies, while preventing ionizing radiation exposure.

Preoperative Lung Ultrasound to Detect Pleural Adhesions: A Systematic Review and Meta-Analysis

The usage of lung ultrasound as a preoperative examination for thoracic surgeries remains controversial. Our systematic review and meta-analysis aimed to evaluate preoperative lung ultrasound diagnostic accuracy for detecting pleural adhesions. We searched articles published in MEDLINE, Embase, CENTRAL, and the International Clinical Trials Registry Platform until October 2020. Inclusion criteria were observational studies, case-control studies, and case series assessing preoperative lung ultrasound diagnostic accuracy. The study quality of included articles was evaluated using the modified quality assessment of diagnostic accuracy studies-2 tool. The pooled sensitivity and specificity were calculated using the bivariate random-effects model. The overall quality of evidence was summarized using the grading of recommendations, assessment, development, and evaluation approach. Eleven articles were included in our systematic review. A high risk of bias was noted regarding undefined pleural adhesions and non-predefined pathological diagnosis. Based on the ten articles included for meta-analysis, the pooled sensitivity and specificity were 71% [95% confidence interval (CI), 56%-82%], and 96% (95% CI, 89%-99%), respectively. The overall quality of evidence was moderate. Our systematic review revealed that lung ultrasound had high specificity. It may serve as a rule-in test for detecting pleural adhesions before thoracic surgeries, which may assist surgeons in preparation for a prolonged surgery or increased risk of complications that occurred by trocar insertion such as bleeding and persistent air leak.

Speckle tracking quantification of lung sliding for the diagnosis of pneumothorax: a multicentric observational study

PURPOSE

Lung ultrasound is used for the diagnosis of pneumothorax, based on lung sliding abolition which is a qualitative and operator-dependent assessment. Speckle tracking allows the quantification of structure deformation over time by analysing acoustic markers. We aimed to test the ability of speckle tracking technology to quantify lung sliding in a selected cohort of patients and to observe how the technology may help the process of pneumothorax diagnosis.

METHODS

We performed retrospectively a pleural speckle tracking analysis on ultrasound loops from patients with pneumothorax. We compared the values measured by two observers from pneumothorax side with contralateral normal lung side. The receiver operating characteristic (ROC) curve was constructed to evaluate the performance of maximal pleural strain to detect the lung sliding abolition. Diagnosis performance and time to diagnosis between B-Mode and speckle tracking technology were compared from a third blinded observer.

RESULTS

We analysed 104 ultrasound loops from 52 patients. The area under the ROC curve of the maximal pleural strain value to identify lung sliding abolition was 1.00 [95%CI 1.00; 1.00]. Specificity was 100% [95%CI 93%; 100%] and sensitivity was 100% [95%CI 93%; 100%] with the best cut-off of 4%. Over 104 ultrasound loops, the blinded observer made two errors with B-Mode and none with speckle tracking. The median diagnosis time was 3 [2-5] seconds for B-Mode versus 2 [1-2] seconds for speckle tracking (p = 0.001).

CONCLUSION

Speckle tracking technology allows lung sliding quantification and detection of lung sliding abolition in case of pneumothorax on selected ultrasound loops.

Diagnostic accuracy of a novel software technology for detecting pneumothorax in a porcine model

INTRODUCTION

Our objective was to measure the diagnostic accuracy of a novel software technology to detect pneumothorax on Brightness (B) mode and Motion (M) mode ultrasonography.

METHODS

Ultrasonography fellowship-trained emergency physicians performed thoracic ultrasonography at baseline and after surgically creating a pneumothorax in eight intubated, spontaneously breathing porcine subjects. Prior to pneumothorax induction, we captured sagittal M-mode still images and B-mode videos of each intercostal space with a linear array transducer at 4cm of depth. After collection of baseline images, we placed a chest tube, injected air into the pleural space in 250mL increments, and repeated the ultrasonography for pneumothorax volumes of 250mL, 500mL, 750mL, and 1000mL. We confirmed pneumothorax with intrapleural digital manometry and ultrasound by expert sonographers. We exported collected images for interpretation by the software. We treated each individual scan as a single test for interpretation by the software.

RESULTS

Excluding indeterminate results, we collected 338M-mode images for which the software demonstrated a sensitivity of 98% (95% confidence interval [CI] 92-99%), specificity of 95% (95% CI 86-99), positive likelihood ratio (LR+) of 21.6 (95% CI 7.1-65), and negative likelihood ratio (LR-) of 0.02 (95% CI 0.008-0.046). Among 364 B-mode videos, the software demonstrated a sensitivity of 86% (95% CI 81-90%), specificity of 85% (81-91%), LR+ of 5.7 (95% CI 3.2-10.2), and LR- of 0.17 (95% CI 0.12-0.22).

CONCLUSIONS

This novel technology has potential as a useful adjunct to diagnose pneumothorax on thoracic ultrasonography.

Thoracic Ultrasound: What Non-radiologists Need to Know

Purpose of review

The aim of this review is to provide the theoretical and practical knowledge essential for non-radiologists to develop the skills necessary to apply thoracic ultrasound as an extension of clinical assessment and intervention.

Recent findings

Issues relating to training and competence are discussed and a library of thoracic ultrasound videos is provided to illustrate artefacts, pleural, parenchymal and pneumothorax pathology as well as important pitfalls to consider. Novel and future diagnostic applications of thoracic ultrasound in the setting of acute cardiorespiratory pathology including consolidation, acute interstitial syndromes and pulmonary embolism are explored.

Summary

Thoracic ultrasound requires an understanding of imaging artefact specific to lung and pleura and a working knowledge of machine knobology for image optimisation and interpretation. Ultrasound is a valuable tool for the practicing chest clinician providing diagnostic information for the assessment of pleural and parenchymal disease and increased safety and cost effectiveness of thoracic interventions.

Electronic supplementary material

The online version of this article (doi:10.1007/s13665-017-0164-1) contains supplementary material, which is available to authorized users.