Prospective Real-Time Validation of a Lung Ultrasound Deep Learning Model in the ICU.

OBJECTIVES: To evaluate the accuracy of a bedside, real-time deployment of a deep learning (DL) model capable of distinguishing between normal (A line pattern) and abnormal (B line pattern) lung parenchyma on lung ultrasound (LUS) in critically ill patients. DESIGN: Prospective, observational study evaluating the performance of a previously trained LUS DL model. Enrolled patients received a LUS examination with simultaneous DL model predictions using a portable device. Clip-level model predictions were analyzed and compared with blinded expert review for A versus B line pattern. Four prediction thresholding approaches were applied to maximize model sensitivity and specificity at bedside. SETTING: Academic ICU. PATIENTS: One-hundred critically ill patients admitted to ICU, receiving oxygen therapy, and eligible for respiratory imaging were included. Patients who were unstable or could not undergo an LUS examination were excluded. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: A total of 100 unique ICU patients (400 clips) were enrolled from two tertiary-care sites. Fifty-six patients were mechanically ventilated. When compared with gold standard expert annotation, the real-time inference yielded an accuracy of 95%, sensitivity of 93%, and specificity of 96% for identification of the B line pattern. Varying prediction thresholds showed that real-time modification of sensitivity and specificity according to clinical priorities is possible. CONCLUSIONS: A previously validated DL classification model performs equally well in real-time at the bedside when platformed on a portable device. As the first study to test the feasibility and performance of a DL classification model for LUS in a dedicated ICU environment, our results justify further inquiry into the impact of employing real-time automation of medical imaging into the care of the critically ill.

Enhancing Annotation Efficiency with Machine Learning: Automated Partitioning of a Lung Ultrasound Dataset by View.

BACKGROUND: Annotating large medical imaging datasets is an arduous and expensive task, especially when the datasets in question are not organized according to deep learning goals. Here, we propose a method that exploits the hierarchical organization of annotating tasks to optimize efficiency. METHODS: We trained a machine learning model to accurately distinguish between one of two classes of lung ultrasound (LUS) views using 2908 clips from a larger dataset. Partitioning the remaining dataset by view would reduce downstream labelling efforts by enabling annotators to focus on annotating pathological features specific to each view. RESULTS: In a sample view-specific annotation task, we found that automatically partitioning a 780-clip dataset by view saved 42 min of manual annotation time and resulted in 55±6 additional relevant labels per hour. CONCLUSIONS: Automatic partitioning of a LUS dataset by view significantly increases annotator efficiency, resulting in higher throughput relevant to the annotating task at hand. The strategy described in this work can be applied to other hierarchical annotation schemes.

Accurate assessment of the lung sliding artefact on lung ultrasonography using a deep learning approach.

Pneumothorax is a potentially life-threatening condition that can be rapidly and accurately assessed via the lung sliding artefact generated using lung ultrasound (LUS). Access to LUS is challenged by user dependence and shortage of training. Image classification using deep learning methods can automate interpretation in LUS and has not been thoroughly studied for lung sliding. Using a labelled LUS dataset from 2 academic hospitals, clinical B-mode (also known as brightness or two-dimensional mode) videos featuring both presence and absence of lung sliding were transformed into motion (M) mode images. These images were subsequently used to train a deep neural network binary classifier that was evaluated using a holdout set comprising 15% of the total data. Grad-CAM explanations were examined. Our binary classifier using the EfficientNetB0 architecture was trained using 2535 LUS clips from 614 patients. When evaluated on a test set of data uninvolved in training (540 clips from 124 patients), the model performed with a sensitivity of 93.5%, specificity of 87.3% and an area under the receiver operating characteristic curve (AUC) of 0.973. Grad-CAM explanations confirmed the model's focus on relevant regions on M-mode images. Our solution accurately distinguishes between the presence and absence of lung sliding artefacts on LUS.

Automation of Lung Ultrasound Interpretation via Deep Learning for the Classification of Normal versus Abnormal Lung Parenchyma: A Multicenter Study.

Lung ultrasound (LUS) is an accurate thoracic imaging technique distinguished by its handheld size, low-cost, and lack of radiation. User dependence and poor access to training have limited the impact and dissemination of LUS outside of acute care hospital environments. Automated interpretation of LUS using deep learning can overcome these barriers by increasing accuracy while allowing point-of-care use by non-experts. In this multicenter study, we seek to automate the clinically vital distinction between A line (normal parenchyma) and B line (abnormal parenchyma) on LUS by training a customized neural network using 272,891 labelled LUS images. After external validation on 23,393 frames, pragmatic clinical application at the clip level was performed on 1162 videos. The trained classifier demonstrated an area under the receiver operating curve (AUC) of 0.96 (±0.02) through 10-fold cross-validation on local frames and an AUC of 0.93 on the external validation dataset. Clip-level inference yielded sensitivities and specificities of 90% and 92% (local) and 83% and 82% (external), respectively, for detecting the B line pattern. This study demonstrates accurate deep-learning-enabled LUS interpretation between normal and abnormal lung parenchyma on ultrasound frames while rendering diagnostically important sensitivity and specificity at the video clip level.

Pandemic infection rates are deterministic but cannot be modeled.

The covid-19 infection rates for a large number of infections collected from a large number of different sites are well defined with a negligible scatter. The simplest invertible iterated map, exponential growth and decay, emerges from country-wide histograms whenever Tchebychev's inequality is satisfied to within several decimal places. This is one point. Another is that failed covid-19 pandemic model predictions have been reported repeatedly by the news media. Model predictions fail because the observed infection rates are beyond modeling: any model that uses fixed rates or uses memory or averages of past rates cannot reproduce the data on active infections. When those possibilities are ruled out, then little is left. Under lockdown and social distancing, the rates unfold daily in small but unforeseeable steps, they are algorithmically complex. We can, however, use two days in the daily data, today and any single day in the past (generally yesterday), to make a useful forecast of future infections. No model provides results better than this simple forecast. We analyze the actual doubling times for covid-19 data and compare them with our predicted doubling times. Flattening and peaking are precisely defined. We identify and study the separate effects of social distancing vs recoveries in the daily infection rates. Social distancing can only cause flattening but recoveries are required in order for the active infections to peak and decay. Three models and their predictions are analyzed. Pandemic data for Austria, Germany, Italy, the USA, the UK, Finland, China, Taiwan, and Sweden are discussed.

Anomalous scaling of stochastic processes and the Moses effect.

The state of a stochastic process evolving over a time t is typically assumed to lie on a normal distribution whose width scales like t^{1/2}. However, processes in which the probability distribution is not normal and the scaling exponent differs from 1/2 are known. The search for possible origins of such "anomalous" scaling and approaches to quantify them are the motivations for the work reported here. In processes with stationary increments, where the stochastic process is time-independent, autocorrelations between increments and infinite variance of increments can cause anomalous scaling. These sources have been referred to as the Joseph effect and the Noah effect, respectively. If the increments are nonstationary, then scaling of increments with t can also lead to anomalous scaling, a mechanism we refer to as the Moses effect. Scaling exponents quantifying the three effects are defined and related to the Hurst exponent that characterizes the overall scaling of the stochastic process. Methods of time series analysis that enable accurate independent measurement of each exponent are presented. Simple stochastic processes are used to illustrate each effect. Intraday financial time series data are analyzed, revealing that their anomalous scaling is due only to the Moses effect. In the context of financial market data, we reiterate that the Joseph exponent, not the Hurst exponent, is the appropriate measure to test the efficient market hypothesis.

Evaluation of 39 medical implants at 7.0 T.

OBJECTIVE: With increased signal to noise ratios, 7.0-T MRI has the potential to contribute unique information regarding anatomy and pathophysiology of a disease. However, concerns for the safety of subjects with metallic medical implants have hindered advancement in this field. The purpose of the present research was to evaluate the MRI safety for 39 commonly used medical implants at 7.0 T. METHODS: Selected metallic implants were tested for magnetic field interactions, radiofrequency-induced heating and artefacts using standardized testing techniques. RESULTS: 5 of the 39 implants tested may be unsafe for subjects undergoing MRI at 7.0 T. CONCLUSION: Implants were deemed either "MR Conditional" or "MR Unsafe" for the 7.0-T MRI environment. Further research is needed to expand the existing database categorizing implants that are acceptable for patients referred for MRI examinations at 7.0 T. ADVANCES IN KNOWLEDGE: Lack of MRI testing for common metallic medical implants limits the translational potential of 7.0-T MRI. For safety reasons, patients with metallic implants are not allowed to undergo a 7.0-T MRI scan, precluding part of the population that can benefit from the detailed resolution of ultra-high-field MRIs. This investigation provides necessary MRI testing of common medical implants at 7.0 T.

Nonstationary increments, scaling distributions, and variable diffusion processes in financial markets.

Fat-tailed distributions have been reported in fluctuations of financial markets for more than a decade. Sliding interval techniques used in these studies implicitly assume that the underlying stochastic process has stationary increments. Through an analysis of intraday increments, we explicitly show that this assumption is invalid for the Euro-Dollar exchange rate. We find several time intervals during the day where the standard deviation of increments exhibits power law behavior in time. Stochastic dynamics during these intervals is shown to be given by diffusion processes with a diffusion coefficient that depends on time and the exchange rate. We introduce methods to evaluate the dynamical scaling index and the scaling function empirically. In general, the scaling index is significantly smaller than previously reported values close to 0.5. We show how the latter as well as apparent fat-tailed distributions can occur only as artifacts of the sliding interval analysis.