Make It Less Complex: Autoencoder for Speckle Noise Removal-Application to Breast and Lung Ultrasound.

Ultrasound (US) imaging is used in the diagnosis and monitoring of COVID-19 and breast cancer. The presence of Speckle Noise (SN) is a downside to its usage since it decreases lesion conspicuity. Filters can be used to remove SN, but they involve time-consuming computation and parameter tuning. Several researchers have been developing complex Deep Learning (DL) models (150,000-500,000 parameters) for the removal of simulated added SN, without focusing on the real-world application of removing naturally occurring SN from original US images. Here, a simpler (<30,000 parameters) Convolutional Neural Network Autoencoder (CNN-AE) to remove SN from US images of the breast and lung is proposed. In order to do so, simulated SN was added to such US images, considering four different noise levels (σ = 0.05, 0.1, 0.2, 0.5). The original US images (N = 1227, breast + lung) were given as targets, while the noised US images served as the input. The Structural Similarity Index Measure (SSIM) and Peak Signal-to-Noise Ratio (PSNR) were used to compare the output of the CNN-AE and of the Median and Lee filters with the original US images. The CNN-AE outperformed the use of these classic filters for every noise level. To see how well the model removed naturally occurring SN from the original US images and to test its real-world applicability, a CNN model that differentiates malignant from benign breast lesions was developed. Several inputs were used to train the model (original, CNN-AE denoised, filter denoised, and noised US images). The use of the original US images resulted in the highest Matthews Correlation Coefficient (MCC) and accuracy values, while for sensitivity and negative predicted values, the CNN-AE-denoised US images (for higher σ values) achieved the best results. Our results demonstrate that the application of a simpler DL model for SN removal results in fewer misclassifications of malignant breast lesions in comparison to the use of original US images and the application of the Median filter. This shows that the use of a less-complex model and the focus on clinical practice applicability are relevant and should be considered in future studies.

Influence of Ultrasound Settings on Laboratory Vertical Artifacts.

OBJECTIVE: The aim of the work described here was to analyze the relationship between the change in ultrasound (US) settings and the vertical artifacts' number, visual rating and signal intensity METHODS: An in vitro phantom consisting of a damp sponge and gelatin mix was created to simulate vertical artifacts. Furthermore, several US parameters were changed sequentially (i.e., frequency, dynamic range, line density, gain, power and image enhancement) and after image acquisition. Five US experts rated the artifacts for number and quality. In addition, a vertical artifact visual score was created to determine the higher artifact rating ("optimal") and the lower artifact rating ("suboptimal"). Comparisons were made between the tested US parameters and baseline recordings. RESULTS: The expert intraclass correlation coefficient for the number of vertical artifacts was 0.694. The parameters had little effect on the "optimal" vertical artifacts but changed their number. Dynamic range increased the number of discernible vertical artifacts to 3 from 36 to 102 dB. CONCLUSION: The intensity did not correlate with the visual rating score. Most of the available US parameters did not influence vertical artifacts.

Machine learning for the real-time assessment of left ventricular ejection fraction in critically ill patients: a bedside evaluation by novices and experts in echocardiography.

BACKGROUND: Machine learning algorithms have recently been developed to enable the automatic and real-time echocardiographic assessment of left ventricular ejection fraction (LVEF) and have not been evaluated in critically ill patients. METHODS: Real-time LVEF was prospectively measured in 95 ICU patients with a machine learning algorithm installed on a cart-based ultrasound system. Real-time measurements taken by novices (LVEFNov) and by experts (LVEFExp) were compared with LVEF reference measurements (LVEFRef) taken manually by echo experts. RESULTS: LVEFRef ranged from 26 to 80% (mean 54 ± 12%), and the reproducibility of measurements was 9 ± 6%. Thirty patients (32%) had a LVEFRef < 50% (left ventricular systolic dysfunction). Real-time LVEFExp and LVEFNov measurements ranged from 31 to 68% (mean 54 ± 10%) and from 28 to 70% (mean 54 ± 9%), respectively. The reproducibility of measurements was comparable for LVEFExp (5 ± 4%) and for LVEFNov (6 ± 5%) and significantly better than for reference measurements (p < 0.001). We observed a strong relationship between LVEFRef and both real-time LVEFExp (r = 0.86, p < 0.001) and LVEFNov (r = 0.81, p < 0.001). The average difference (bias) between real time and reference measurements was 0 ± 6% for LVEFExp and 0 ± 7% for LVEFNov. The sensitivity to detect systolic dysfunction was 70% for real-time LVEFExp and 73% for LVEFNov. The specificity to detect systolic dysfunction was 98% both for LVEFExp and LVEFNov. CONCLUSION: Machine learning-enabled real-time measurements of LVEF were strongly correlated with manual measurements obtained by experts. The accuracy of real-time LVEF measurements was excellent, and the precision was fair. The reproducibility of LVEF measurements was better with the machine learning system. The specificity to detect left ventricular dysfunction was excellent both for experts and for novices, whereas the sensitivity could be improved. TRIAL REGISTRATION: NCT05336448. Retrospectively registered on April 19, 2022.

Time course of lung ultrasound findings in patients with COVID-19 pneumonia and cardiac dysfunction.

BACKGROUND: Lung ultrasound (LUS) is a valuable tool to predict and monitor the COVID-19 pneumonia course. However, the influence of cardiac dysfunction (CD) on LUS findings remains to be studied. Our objective was to determine the effect of CD on LUS in hospitalized patients with COVID-19 pneumonia. MATERIAL AND METHODS: Fifty-one patients with COVID-19 pneumonia participated in the study. Focused echocardiography (FoCUS) was carried out on day 1 to separate patients into two groups depending on whether they had FoCUS signs of CD (CD+ vs CD-). LUS scores, based on the thickness of the pleural line, the B-line characteristics, and the presence or not of consolidations, were obtained three times along the patient's admission (D1, D5, D10) and compared between CD+ and CD- patients. A correlation analysis was carried out between LUS scores and the ratio of the arterial partial pressure of oxygen to the fraction of the inspired oxygen (P/F ratio). RESULTS: Twenty-two patients were CD+ and 29 patients were CD-. Among the CD+ patients, 19 were admitted to the intensive care unit (ICU), seven received invasive mechanical ventilation (IMV), and one did not survive. Among the CD- patients, 11 were admitted to the ICU, one received IMV and seven did not survive. CD+ patients showed a significantly lower P/F ratio than CD- patients. However, LUS scores showed no between-group differences, except for fewer subpleural consolidations in the upper quadrants of CD+ than on CD- patients. CONCLUSION: In patients with COVID-19, CD contributed to a worse clinical course, but it did not induce significant changes in LUS. Our findings suggest that pathophysiological factors other than those reflected by LUS may be responsible for the differences in clinical condition between CD+ and CD- patients.

Impact of Navigated Task-specific fMRI on Direct Cortical Stimulation.

BACKGROUND AND STUDY AIMS: Cortical mapping (CM) with direct cortical stimulation (DCS) in awake craniotomy is used to preserve cognitive functions such as language. Nevertheless, patient collaboration during this procedure is influenced by previous neurological symptoms and growing discomfort with DCS duration. Our study aimed to evaluate the impact of navigated task-specific functional magnetic resonance imaging (nfMRI) on the practical aspects of DCS. MATERIAL AND METHODS: We recruited glioma patients scheduled for awake craniotomy for prior fMRI-based CM, acquired during motor and language tasks (i.e., verb generation, semantic and syntactic decision tasks). Language data was combined to generate a probabilistic map indicating brain regions activated with more than one paradigm. Presurgical neurophysiological language tests (i.e., verb generation, picture naming, and semantic tasks) were also performed. We considered for subsequent study only the patients with a minimum rate of correct responses of 50% in all tests. These patients were then randomized to perform intraoperative language CM either using the multimodal approach (mCM), using nfMRI and DCS combined, or electrical CM (eCM), with DCS alone. DCS was done while the patient performed picture naming and nonverbal semantic decision tasks. Methodological features such as DCS duration, number of stimuli, total delivered stimulus duration per task, and frequency of seizures were analyzed and compared between groups. The correspondence between positive responses obtained with DCS and nfMRI was also evaluated. RESULTS: Twenty-one surgeries were included, thirteen of which using mCM (i.e., test group). Patients with lower presurgical neuropsychological performance (correct response rate between 50 and 80% in language tests) showed a decreased DCS duration in comparison with the control group. None of the compared methodological features showed differences between groups. Correspondence between DCS and nfMRI was 100/84% in the identification of the precentral gyrus for motor function/opercular frontal inferior gyrus for language function, respectively. CONCLUSION: Navigated fMRI data did not influence DCS in practice. Presurgical language disturbances limited the applicability of DCS mapping in awake surgery.

Potencial de ação duplo e alterações ecográficas do nervo mediano em doentes com síndrome do túnel cárpico / Double peak action potential and ultrasonography nerve evaluation in patients with carpal tunnel syndrome

Introdução: A síndrome do túnel cárpico (STC) é sustentada pela presença de lentificação do potencial de ação do nervo mediano no estudo ortodrómico de condução nervosa sensitiva em comparação com o nervo radial (EC-MR). Uma técnica modificada de registo simultâneo de ambos os nervos (EC-MRsimul) pode ser utilizada, detetando a lentificação do nervo mediano através da presença de potencial de ação sensitivo duplo (PAS-D). O objetivo deste estudo é correlacionar o PAS-D com os achados de EC-MR e com o edema perineural em doentes com diagnóstico de STC. Métodos: Indivíduos saudáveis e doentes com STC há menos de 12 meses foram submetidos a avaliação eletrofisiológica, incluindo o EC-MR e o EC-MRsimul. Foi também realizada ultrassonografia para registo da área seccional do nervo mediano (AS-NM) no punho e antebraço e, respetivo índice punho-antebraço (I-PA). Resultados: Foram recrutados 38 doentes com idade média de 54,8 ± 15,3 anos com STC e 18 indivíduos saudáveis. A diferença de latência distal entre o nervo mediano e radial foi superior nos doentes (0,80 ± 0,30ms vs. 0,15 ± 0,20ms; p=0,015). O EC­MRsimul demonstrou a presença de PAS-D nos indivíduos sintomáticos. A AS­NM no punho foi também superior nos doentes (8,9 ± 0,9mm2 vs. 6,6 ± 0,7mm2; p<0.003), tendo sido identificado um quisto sinovial e um neurinoma. O edema perineural traduzido pela AS-SM correlacionou-se positivamente com a diferença de latência interpico no EC-MR e com a presença de PAS-D. Conclusão: O PAS-D está associado à presença de lentificação da condução e ao edema perineural do nervo mediano no contexto de STC ou devido a lesões ocupantes de espaço. (AU)

Introduction: Carpal tunnel syndrome (CTS) is characterized by median nerve action potential slowing, which can be shown in comparative orthodromic sensory nerve conduction studies between median and radial nerve (NCS-MR). A modified technique with simultaneous recording of both nerves (NCS-MRsimul) can also be used to detect median nerve slowing through the presence of double peak action potential (DPp). The study aims to correlate the presence of DPp with NCS-MR findings and with perineural edema in patients diagnosed with STC. Methods: Healthy individuals and patients with CTS for less than 12 months underwent NCS-MR and MRsimul. An ultrasonography evaluation was also performed to record the medial nerve sectional area (SA-MN) on the wrist and forearm, and the wrist-to-forearm ratio (WFR) was calculated. Results: We recruited 38 patients with CTS whose mean age was 54.8 ± 15.3 years and 18 healthy individuals. Distal latency difference between the median and radial nerves was higher in patients with CTS (0.80 ± 0.30 ms vs. 0.15 ± 0.20 ms, p = 0.015). NCS-MRsimul showed DPp in symptomatic individuals. SA-MN in the wrist was also higher in patients with CTS (8.9 ± 0.9 mm2 vs. 6.6 ± 0.7 mm2 , p <0.003). Ultrasonography evaluation identified a synovial cyst and a neurinoma. Perineural edema traduced by higher SA-MN and WFR correlated positively with interpeak latency difference in NCS-MR and with the presence of DPp. Conclusions: DPp was associated with median nerve sensory action potential slowing and with perineural edema due to either CTS or to space-occupying lesions. (AU)

Reconstruction of white matter fibre tracts using diffusion kurtosis tensor imaging at 1.5T: Pre-surgical planning in patients with gliomas.

Tractography studies for pre-surgical planning of primary brain tumors is typically done using diffusion tensor imaging (DTI), which cannot resolve crossing, kissing or highly angulated fibres. Tractography based on the estimation of the diffusion kurtosis (DK) tensor was recently demonstrated to enable tackling these limitations. However, its use in the clinical context at low 1.5T field has not yet been reported. PURPOSE: To evaluate if the estimation of whole-brain tractography using the DK tensor is feasible for pre-surgical investigation of patients with brain tumors at 1.5T. METHODS: Eight healthy subjects and 3 patients with brain tumors were scanned at 1.5T using a 12-channel head coil. Diffusion-weighted images were acquired with repetition/echo times of 5800/107 ms, 82 × 82 resolution, 3 × 3 × 3 mm3 voxel size, b-values of 0, 1000, 2000 s/mm2 and 64 gradient sensitising directions. Whole-brain tractography was estimated using the DK tensor and corticospinal tracts (CST) were isolated using regions-of-interest placed at the cerebral peduncles and motor gyrus. Tract size, DK metrics and CST deviation index (highest curvature point) were compared between healthy subjects and patients. RESULTS: Tract sizes did not differ between groups. The CST deviation index was significantly higher in patients compared to healthy subjects. Fractional anisotropy was significantly lower in patients, with higher mean kurtosis asymmetry index at the highest curvature point in patients. CONCLUSIONS: Corticospinal fibre bundles estimated using DK tensor in a 1.5T scanner presented similar properties in patients with brain gliomas as those reported in the literature using DTI-based tractography.

Motor preparation in picture naming tasks.

In certain circumstances, words can be uttered as an involuntary action. We hypothesize that, once pronunciation of a word is fully prepared it can be triggered as a reflex with no need for cortical processing. We used modified protocols of picture naming tasks, with different levels of cognitive demands, to measure reaction time to word pronunciation (RTWP). In test trials, picture presentation was accompanied by a startling auditory stimulus (SAS). When one and the same picture was repeatedly shown, SAS shortened RTWP by about 30% (StartReact effect), which did not occur when random pictures were shown. If subjects were led to learn which picture was to appear after repeated presentation of three pictures in sequence, they exhibited again the StartReact effect. We conclude that word pronunciation may be fully prepared for execution in absence of cognitive demands. However, the StartReact effect is inhibited during cognitive tasks.

Double peak sensory nerve action potentials to single stimuli in nerve conduction studies.

INTRODUCTION: In humans, sensory nerve action potentials (SNAPs) can show 2 separate deflections, i.e., double peak potentials (DPp), which necessarily means that 1 peak is delayed with respect to the other. DPps may have various origins and be due to either physical or physiological properties. METHODS: We review the nature of commonly encountered DPps in clinical practice, provide the most likely interpretations for their physiological origin, and assess their reproducibility and clinical utility. RESULTS: We classified the DPps into 3 categories: (1) simultaneous anodal and cathodal stimulation. (2) simultaneous recording from 2 different nerves at the same site, and (3) SNAP desynchronization. CONCLUSIONS: Although the recording of DPps is not a standardized neurophysiological method, their study brings interesting cues about the physiology of nerve stimulation and paves the way for clinical application of such an observation. Muscle Nerve 55: 619-625, 2017.

Stimulus waveform determines the characteristics of sensory nerve action potentials.

OBJECTIVE: In routine nerve conduction studies supramaximal electrical stimuli generate sensory nerve action potentials by depolarization of nerve fibers under the cathode. However, stimuli of submaximal intensity may give rise to action potentials generated under the anode. We tested if this phenomenon depends on the characteristics of stimulus ending. METHODS: We added a circuit to our stimulation device that allowed us to modify the end of the stimulus by increasing the time constant of the decay phase. RESULTS: Increasing the fall time caused a reduction of anode action potential (anAP) amplitude, and eventually abolished it, in all tested subjects. We subsequently examined the stimulus waveform in a series of available electromyographs stimulators and found that the anAP could only be obtained with stimulators that issued stimuli ending sharply. CONCLUSION: Our results prove that the anAP is generated at stimulus end, and depends on the sharpness of current shut down. Electromyographs produce stimuli of varying characteristics, which limits the reproducibility of anAP results by interested researchers. SIGNIFICANCE: The study of anodal action potentials might be a useful tool to have a quick appraisal of distal human sensory nerve excitability.

Anodal sensory nerve action potentials: From physiological understanding to potential clinical applicability.

INTRODUCTION: Low-intensity electrical stimuli of digital nerves may generate a double peak potential (DPp), composed of a cathodal (caAP) and an anodal (anAP) potential in orthodromic recordings. METHODS: We studied the effects on caAP and anAP of stimuli of variable intensity, duration, and frequency. We also applied a conditioning stimulus to study potential differences in recovery time. RESULTS: The anAP was obtained in 33 of 40 healthy subjects (82.5%) and 4 of 20 patients with various types of sensory neuropathies (20%). Changes in stimulus duration and intensity had reciprocal effects on the amplitude of the anAP and the caAP. There were significant differences in recovery time between caAP and anAP after a conditioning stimulus. CONCLUSION: The caAP and anAP are 2 interdependent waveforms generated by different effects of the same stimulus over axons at the verge of depolarization. Muscle Nerve 53: 897-905, 2016.

Antidromic vs orthodromic sensory median nerve conduction studies.

OBJECTIVE: Median sensory nerve conduction studies are arguably the most often performed electrodiagnostic tests worldwide. Routine tests in clinical practice are done using either antidromic or orthodromic techniques type of stimulation, with no universal agreement on the use of one or the other technique. METHODS: We review the advantages and drawbacks of antidromic and orthodromic as well as their particularities for clinical application and research. RESULTS: The two techniques differ on how physical and physiological changes affect the action potential. Near-nerve recording is better suited for the orthodromic than for the antidromic technique, while studies of nerve excitability are better suited for the antidromic than for the orthodromic technique. CONCLUSION: Both techniques are equally suitable for routine tests but research studies may specifically demand one or the other.