Automatic Optic Nerve Measurement: A New Tool to Standardize Optic Nerve Assessment in Ultrasound B-Mode Images

Automation of ultrasonographic optic nerve sheath diameter measurement using convolutional neural networks

BACKGROUND AND PURPOSE

Ultrasonographic optic nerve sheath (ONS) diameter is a noninvasive intracranial pressure (ICP) surrogate. ICP is monitored invasively in specialized intensive care units. Noninvasive ICP monitoring is important in less specialized settings. However, noninvasive ICP monitoring using ONS diameter (ONSD) is limited by the need for experts to obtain and perform measurements. We aim to automate ONSD measurements using a deep convolutional neural network (CNN) with a novel masking technique.

METHODS

We trained a CNN to reproduce masks that mark the ONS. The edges of the mask are defined by an expert. Eight models were trained with 1000 epochs per model. The Dice-similarity-coefficient-weighted averaged outputs of the eight models yielded the final predicted mask. Eight hundred and seventy-three images were obtained from 52 transorbital cine-ultrasonography sessions, performed on 46 patients with brain injuries. Eight hundred and fourteen images from 48 scanning sessions were used for training and validation and 59 images from four sessions for testing. Bland-Altman and Pearson linear correlation analyses were used to evaluate the agreement between CNN and expert measurements.

RESULTS

Expert ONSD measurements and CNN-derived ONSD estimates had strong agreement (r = 0.7, p < .0001). The expert mean ONSD (standard deviation) is 5.27 mm (0.43) compared to CNN mean estimate of 5.46 mm (0.37). Mean difference (95% confidence interval, p value) is 0.19 mm (0.10-0.27 mm, p = .0011), and root mean square error is 0.27 mm.

CONCLUSION

A CNN can learn ONSD measurement using masking without image segmentation or landmark detection.

Effects of head-down tilt on optic nerve sheath diameter in healthy subjects

PURPOSE

Intracranial pressure increases in head-down tilt (HDT) body posture. This study evaluated the effect of HDT on the optic nerve sheath diameter (ONSD) in normal subjects.

METHODS

Twenty six healthy adults (age 28 [4.7] years) participated in seated and 6° HDT visits. For each visit, subjects presented at 11:00 h for baseline seated scans and then maintained a seated or 6° HDT posture from 12:00 to 15:00 h. Three horizontal axial and three vertical axial scans were obtained at 11:00, 12:00 and 15:00 h with a 10 MHz ultrasonography probe on the same eye, randomly chosen per subject. At each time point, horizontal and vertical ONSD (mm) were quantified by averaging three measures taken 3 mm behind the globe.

RESULTS

In the seated visit, ONSDs were similar across time (p > 0.05), with an overall mean (standard deviation) of 4.71 (0.48) horizontally and 5.08 (0.44) vertically. ONSD was larger vertically than horizontally at each time point (p < 0.001). In the HDT visit, ONSD was significantly enlarged from baseline at 12:00 and 15:00 h (p < 0.001 horizontal and p < 0.05 vertical). Mean (standard error) horizontal ONSD change from baseline was 0.37 (0.07) HDT versus 0.10 (0.05) seated at 12:00 h (p = 0.002) and 0.41 (0.09) HDT versus 0.12 (0.06) seated at 15:00 h (p = 0.002); mean vertical ONSD change was 0.14 (0.07) HDT versus -0.07 (0.04) seated at 12:00 h (p = 0.02) and 0.19 (0.06) HDT versus -0.03 (0.04) seated at 15:00 h (p = 0.01). ONSD change in HDT was similar between 12:00 and 15:00 h (p ≥ 0.30). Changes at 12:00 h correlated with those at 15:00 h for horizontal (r = 0.78, p < 0.001) and vertical ONSD (r = 0.73, p < 0.001).

CONCLUSION

The ONSD increased when body posture transitioned from seated to HDT position without any further change at the end of the 3 h in HDT.

Evaluation of commercially available point-of-care ultrasound for automated optic nerve sheath measurement

BACKGROUND

Measurement of the optic nerve sheath diameter (ONSD) via ultrasonography has been proposed as a non-invasive metric of intracranial pressure that may be employed during in-field patient triage. However, first responders are not typically trained to conduct sonographic exams and/or do not have access to an expensive ultrasound device. Therefore, for successful deployment of ONSD measurement in-field, we believe that first responders must have access to low-cost, portable ultrasound and be assisted by artificial intelligence (AI) systems that can automatically interpret the optic nerve sheath ultrasound scan. We examine the suitability of five commercially available, low-cost, portable ultrasound devices that can be combined with future artificial intelligence algorithms to reduce the training required for and cost of in-field optic nerve sheath diameter measurement. This paper is focused on the quality of the images generated by these low-cost probes. We report results of a clinician preference survey and compare with a lab analysis of three quantitative image quality metrics across devices. We also examine the suitability of the devices in a hypothetical far-forward deployment using operators unskilled in ultrasound, with the assumption of a future onboard AI video interpreter.

RESULTS

We find statistically significant differences in clinician ranking of the devices in the following categories: "Image Quality", "Ease of Acquisition", "Software", and "Overall ONSD". We show differences in signal-to-noise ratio, generalized contrast-to-noise ratio, point-spread function across the devices. These differences in image quality result in a statistically significant difference in manual ONSD measurement. Finally, we show that sufficiently wide transducers can capture the optic nerve sheath during blind (no visible B-mode) scans performed by operators unskilled in sonography.

CONCLUSIONS

Ultrasound of the optic nerve sheath has the potential to be a convenient, non-invasive, point-of-injury or triage measure for elevated intracranial pressure in cases of traumatic brain injury. When transducer width is sufficient, briefly trained operators may obtain video sequences of the optic nerve sheath without guidance. This data suggest that unskilled operators are able to achieve the images needed for AI interpretation. However, we also show that image quality differences between ultrasound probes may influence manual ONSD measurements.

Development of a Deep Learning-Based System for Optic Nerve Characterization in Transorbital Ultrasound Images on a Multicenter Data Set

OBJECTIVE

Characterization of the optic nerve through measurement of optic nerve diameter (OND) and optic nerve sheath diameter (ONSD) using transorbital sonography (TOS) has proven to be a useful tool for the evaluation of intracranial pressure (ICP) and multiple neurological conditions. We describe a deep learning-based system for automatic characterization of the optic nerve from B-mode TOS images by automatic measurement of the OND and ONSD. In addition, we determine how the signal-to-noise ratio in two different areas of the image influences system performance.

METHODS

A UNet was trained as the segmentation model. The training was performed on a multidevice, multicenter data set of 464 TOS images from 110 subjects. Fivefold cross-validation was performed, and the training process was repeated eight times. The final prediction was made as an ensemble of the predictions of the eight single models. Automatic OND and ONSD measurements were compared with the manual measurements taken by an expert with a graphical user interface that mimics a clinical setting.

RESULTS

A Dice score of 0.719 ± 0.139 was obtained on the whole data set merging the test folds. Pearson's correlation was 0.69 for both OND and ONSD parameters. The signal-to-noise ratio was found to influence segmentation performance, but no clear correlation with diameter measurement performance was determined.

CONCLUSION

The developed system has a good correlation with manual measurements, proving that it is feasible to create a model capable of automatically analyzing TOS images from multiple devices. The promising results encourage further definition of a standard protocol for the automatization of the OND and ONSD measurement process using deep learning-based methods. The image data and the manual measurements used in this work will be available at 10.17632/kw8gvp8m8x.1.

Optic nerve sheath diameter and spaceflight: defining shortcomings and future directions

Neuro-ocular changes during long-duration space flight are known as spaceflight-associated neuro-ocular syndrome (SANS). The ability to detect, monitor, and prevent SANS is a priority of current space medicine research efforts. Optic nerve sheath diameter (ONSD) measurement has been used both terrestrially and in microgravity as a proxy for measurements of elevated intracranial pressure. ONSD shows promise as a potential method of identifying and quantitating neuro-ocular changes during space flight. This review examines 13 studies measuring ONSD and its relationship to microgravity exposure or ground-based analogs, including head-down tilt, dry immersion, or animal models. The goal of this correspondence is to describe heterogeneity in the use of ONSD in the current SANS literature and make recommendations to reduce heterogeneity in future studies through standardization of imaging modalities, measurement techniques, and other aspects of study design.

Accuracy of Optic Nerve Sheath Diameter Measurements in Pocket-Sized Ultrasound Devices in a Simulation Model

Introduction

Transorbital sonographic measurement of optic nerve sheath diameter (ONSD) is an emerging non-invasive technique for the identification and monitoring of intracranial hypertension. In recent years, new pocket ultrasound devices have become available, and it is uncertain if they have the resolution to measure such small structures appropriately as compared to their predecessors. In this study, we measure the performance of three ultrasound units on a simulation model to establish their precision and accuracy.

Methods

ONSD was measured by three expert point-of-care sonographers using ultrasound machines three times on each of seven discrete ONS model sizes ranging from 3.5 to 7.9 mm. Two pocket ultrasounds (IVIZ, Sonosite, and Lumify, Philips) and one standard-sized portable ultrasound (M-Turbo, Sonosite) were used. Measurements were analyzed for mean error and variance and tested for significance using blocked covariance matrix regression analyses.

Results

The devices differed in their variances (Lumify: 0.19 mm², M-Turbo: 0.26 mm², IVIZ: 0.34 mm²) and their mean error (Lumify: -0.05 mm, M-Turbo: 0.10 mm, IVIZ: -0.10 mm). The difference in mean error between users is not significant (p = 0.45), but there is a significant difference in mean error between devices (p = 0.02).

Conclusions

Accurate ONSD measurement is possible utilizing pocket-sized ultrasound, and in some cases, may be more accurate than larger portable ultrasound units. While the differences in these devices were statistically significant, all three were highly accurate, with one pocket device (Lumify) outperforming the rest. Further study in human subjects should be conducted prior to using pocket ultrasound devices for in vivo diagnosis of intracranial hypertension.

Nerve optic segmentation in CT images using a deep learning model and a texture descriptor

The increased intracranial pressure (ICP) can be described as an increase in pressure around the brain and can lead to serious health problems. The assessment of ultrasound images is commonly conducted by skilled experts which is a time-consuming approach, but advanced computer-aided diagnosis (CAD) systems can assist the physician to decrease the time of ICP diagnosis. The accurate detection of the nerve optic regions, with drawing a precise slope line behind the eyeball and calculating the diameter of nerve optic, are the main aims of this research. First, the Fuzzy C-mean (FCM) clustering is employed for segmenting the input CT screening images into the different parts. Second, a histogram equalization approach is used for region-based image quality enhancement. Then, the Local Directional Number method (LDN) is used for representing some key information in a new image. Finally, a cascade Convolutional Neural Network (CNN) is employed for nerve optic segmentation by two distinct input images. Comprehensive experiments on the CT screening dataset [The Cancer Imaging Archive (TCIA)] consisting of 1600 images show the competitive results of inaccurate extraction of the brain features. Also, the indexes such as Dice, Specificity, and Precision for the proposed approach are reported 87.7%, 91.3%, and 90.1%, respectively. The final classification results show that the proposed approach effectively and accurately detects the nerve optic and its diameter in comparison with the other methods. Therefore, this method can be used for early diagnose of ICP and preventing the occurrence of serious health problems in patients.

Ultrasound measurement of the optic nerve sheath diameter in traumatic brain injury: a narrative review

Background : Raised intracranial pressure (ICP) needs to be investigated in various situations, especially in traumatic brain injury (TBI). Ultra-sonographic (US) measurement of the optic nerve sheath diameter (ONSD) is a promising noninvasive tool for assessing elevated ICP. Objectives : This narrative review aimed to explain the history of and indications forUS measurement of ONSD. We focused on the detection of elevated ICP after TBI and discussed the possible improvements in detection methods. Conclusions : US measurement of ONSD in TBI cases provides a qualitative but no quantitative assessment of ICP. Current studies usually calculate their own optimum cutoff value for detecting raised ICP based on the balance between sensitivity and specificity of the method when compared with invasive methods. There is no universally accepted threshold. We did not find any paper focusing on the prognosis of patients benefiting from it when compared with usual care. Another limitation is the lack of standardization. US measurement of ONSD cannot be used as the sole technique to detect elevated ICP and monitor its evolution, but it can be a useful tool in a multimodal protocol and it might help to determine the prognosis of patients in various situations.

An Automated Algorithm for Optic Nerve Sheath Diameter Assessment from B-mode Ultrasound Images

BACKGROUND AND PURPOSE

The optic nerve sheath diameter (ONSD) is a promising surrogate marker for the detection of raised intracranial pressure (ICP). However, inconsistencies in manual ONSD assessment are thought to affect ONSD and the corresponding ONSD cutoff values for the diagnosis of elevated ICP, hereby hampering the full potential of ONSD. In this study, we developed an image intensity-invariant algorithm to automatically estimate ONSD from B-mode ultrasound images at multiple depths.

METHODS

The outcomes of the algorithm were validated against manual ONSD measurements by two human experts. Each expert analyzed the images twice (M1 and M2) in unknown order.

RESULTS

The algorithm proved capable of segmenting the ONSD in 39 of 42 images, hereby showing mean differences of -.08 ± .45 and -.05 ± .41 mm compared to averaged ONSD values (M1 + M2/2) of Operator 1 and Operator 2, respectively, whereas the mean difference between the two experts was .03 ± .26 mm. Moreover, differences between algorithm-derived and expert-derived ONSD values were found to be much smaller than the 1 mm difference that is expected between patients with normal and elevated ICP, making it likely that our algorithm can distinguish between these patient groups.

CONCLUSIONS

Our algorithm has the potential to improve the accuracy of ONSD as a surrogate marker for elevated ICP because it has no intrinsic variability. However, future research should be performed to validate if the algorithm does indeed result in more accurate noninvasive ICP predictions.

Hyperosmolar Treatment for Patients at Risk for Increased Intracranial Pressure: A Single-Center Cohort Study

Treatment with osmoactive agents such as mannitol or hypertonic saline (HTS) solutions is widely used to manage or prevent the increase of intracranial pressure (ICP) in central nervous system (CNS) disorders. We sought to evaluate the variability and mean plasma concentrations of the water and electrolyte balance parameters in critically ill patients treated with osmotic therapy and their influence on mortality. This cohort study covered patients hospitalized in an intensive care unit (ICU) from January 2017 to June 2019 with presumed increased ICP or considered to be at risk of it, treated with 15% mannitol (G1, n = 27), a combination of 15% mannitol and 10% hypertonic saline (HTS) (G2, n = 33) or 10% HTS only (G3, n = 13). Coefficients of variation (Cv) and arithmetic means (mean) were calculated for the parameters reflecting the water and electrolyte balance, i.e., sodium (NaCv/NaMean), chloride (ClCv/ClMean) and osmolality (mOsmCv/mOsmMean). In-hospital mortality was also analyzed. The study group comprised 73 individuals (36 men, 49%). Mortality was 67% (n = 49). Median NaCv (G1: p = 0.002, G3: p = 0.03), ClCv (G1: p = 0.02, G3: p = 0.04) and mOsmCv (G1: p = 0.001, G3: p = 0.02) were higher in deceased patients. NaMean (p = 0.004), ClMean (p = 0.04), mOsmMean (p = 0.003) were higher in deceased patients in G3. In G1: NaCv (AUC = 0.929, p < 0.0001), ClCv (AUC = 0.817, p = 0.0005), mOsmCv (AUC = 0.937, p < 0.0001) and in G3: NaMean (AUC = 0.976, p < 0.001), mOsmCv (AUC = 0.881, p = 0.002), mOsmMean (AUC = 1.00, p < 0.001) were the best predictors of mortality. The overall mortality prediction for combined G1+G2+G3 was very good, with AUC = 0.886 (p = 0.0002). The mortality of critically ill patients treated with osmotic agents is high. Electrolyte disequilibrium is the independent predictor of mortality regardless of the treatment method used. Variations of plasma sodium, chloride and osmolality are the most deleterious factors regardless of the absolute values of these parameters.