Lumbar Ultrasound Image Feature Extraction and Classification with Support Vector Machine

Artificial intelligence for ultrasound scanning in regional anaesthesia: a scoping review of the evidence from multiple disciplines

BACKGROUND

Artificial intelligence (AI) for ultrasound scanning in regional anaesthesia is a rapidly developing interdisciplinary field. There is a risk that work could be undertaken in parallel by different elements of the community but with a lack of knowledge transfer between disciplines, leading to repetition and diverging methodologies. This scoping review aimed to identify and map the available literature on the accuracy and utility of AI systems for ultrasound scanning in regional anaesthesia.

METHODS

A literature search was conducted using Medline, Embase, CINAHL, IEEE Xplore, and ACM Digital Library. Clinical trial registries, a registry of doctoral theses, regulatory authority databases, and websites of learned societies in the field were searched. Online commercial sources were also reviewed.

RESULTS

In total, 13,014 sources were identified; 116 were included for full-text review. A marked change in AI techniques was noted in 2016-17, from which point on the predominant technique used was deep learning. Methods of evaluating accuracy are variable, meaning it is impossible to compare the performance of one model with another. Evaluations of utility are more comparable, but predominantly gained from the simulation setting with limited clinical data on efficacy or safety. Study methodology and reporting lack standardisation.

CONCLUSIONS

There is a lack of structure to the evaluation of accuracy and utility of AI for ultrasound scanning in regional anaesthesia, which hinders rigorous appraisal and clinical uptake. A framework for consistent evaluation is needed to inform model evaluation, allow comparison between approaches/models, and facilitate appropriate clinical adoption.

Artificial intelligence and its clinical application in Anesthesiology: a systematic review

PURPOSE

Application of artificial intelligence (AI) in medicine is quickly expanding. Despite the amount of evidence and promising results, a thorough overview of the current state of AI in clinical practice of anesthesiology is needed. Therefore, our study aims to systematically review the application of AI in this context.

METHODS

A systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. We searched Medline and Web of Science for articles published up to November 2022 using terms related with AI and clinical practice of anesthesiology. Articles that involved animals, editorials, reviews and sample size lower than 10 patients were excluded. Characteristics and accuracy measures from each study were extracted.

RESULTS

A total of 46 articles were included in this review. We have grouped them into 4 categories with regard to their clinical applicability: (1) Depth of Anesthesia Monitoring; (2) Image-guided techniques related to Anesthesia; (3) Prediction of events/risks related to Anesthesia; (4) Drug administration control. Each group was analyzed, and the main findings were summarized. Across all fields, the majority of AI methods tested showed superior performance results compared to traditional methods.

CONCLUSION

AI systems are being integrated into anesthesiology clinical practice, enhancing medical professionals' skills of decision-making, diagnostic accuracy, and therapeutic response.

Ultrasound-Based Radiomics for the Classification of Henoch-Schönlein Purpura Nephritis in Children

Henoch-Schönlein purpura nephritis (HSPN) is one of the most common kidney diseases in children. The current diagnosis and classification of HSPN depend on pathological biopsy, which is seriously limited by its invasive and high-risk nature. The aim of the study was to explore the potential of radiomics model for evaluating the histopathological classification of HSPN based on the ultrasound (US) images. A total of 440 patients with Henoch-Schönlein purpura nephritis proved by biopsy were analyzed retrospectively. They were grouped according to two histopathological categories: those without glomerular crescent formation (ISKDC grades I-II) and those with glomerular crescent formation (ISKDC grades III-V). The patients were randomly assigned to either a training cohort (n = 308) or a validation cohort (n = 132) with a ratio of 7:3. The sonologist manually drew the regions of interest (ROI) on the ultrasound images of the right kidney including the cortex and medulla. Then, the ultrasound radiomics features were extracted using the Pyradiomics package. The dimensions of radiomics features were reduced by Spearman correlation coefficients and least absolute shrinkage and selection operator (LASSO) method. Finally, three radiomics models using k-nearest neighbor (KNN), logistic regression (LR), and support vector machine (SVM) were established, respectively. The predictive performance of such classifiers was assessed with receiver operating characteristic (ROC) curve. 105 radiomics features were extracted from derived US images of each patient and 14 features were ultimately selected for the machine learning analysis. Three machine learning models including k-nearest neighbor (KNN), logistic regression (LR), and support vector machine (SVM) were established for HSPN classification. Of the three classifiers, the SVM classifier performed the best in the validation cohort [area under the curve (AUC) =0.870 (95% CI, 0.795-0.944), sensitivity = 0.706, specificity = 0.950]. The US-based radiomics had good predictive value for HSPN classification, which can be served as a noninvasive tool to evaluate the severity of renal pathology and crescentic formation in children with HSPN.

Demystifying Statistics and Machine Learning in Analysis of Structured Tabular Data

Electronic health records have facilitated the extraction and analysis of a vast amount of data with many variables for clinical care and research. Conventional regression-based statistical methods may not capture all the complexities in high-dimensional data analysis. Therefore, researchers are increasingly using machine learning (ML)-based methods to better handle these more challenging datasets for the discovery of hidden patterns in patients' data and for classification and predictive purposes. This article describes commonly used ML methods in structured data analysis with examples in orthopedic surgery. We present practical considerations in starting an ML project and appraising published studies in this field.

Machine Learning Applications in Spine Surgery

This literature review sought to identify and evaluate the current applications of artificial intelligence (AI)/machine learning (ML) in spine surgery that can effectively guide clinical decision-making and surgical planning. By using specific keywords to maximize search sensitivity, a thorough literature research was conducted in several online databases: Scopus, PubMed, and Google Scholar, and the findings were filtered according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A total of 46 studies met the requirements and were included in this review. According to this study, AI/ML models were sufficiently accurate with a mean overall value of 74.9%, and performed best at preoperative patient selection, cost prediction, and length of stay. Performance was also good at predicting functional outcomes and postoperative mortality. Regression analysis was the most frequently utilized application whereas deep learning/artificial neural networks had the highest sensitivity score (81.5%). Despite the relatively brief history of engagement with AI/ML, as evidenced by the fact that 77.5% of studies were published after 2018, the outcomes have been promising. In light of the Big Data era, the increasing prevalence of National Registries, and the wide-ranging applications of AI, such as exemplified by ChatGPT (OpenAI, San Francisco, California), it is highly likely that the field of spine surgery will gradually adopt and integrate AI/ML into its clinical practices. Consequently, it is of great significance for spine surgeons to acquaint themselves with the fundamental principles of AI/ML, as these technologies hold the potential for substantial improvements in overall patient care.

Automatic Detection and Classification of Diabetic Retinopathy Using the Improved Pooling Function in the Convolution Neural Network

Diabetic retinopathy (DR) is an eye disease associated with diabetes that can lead to blindness. Early diagnosis is critical to ensure that patients with diabetes are not affected by blindness. Deep learning plays an important role in diagnosing diabetes, reducing the human effort to diagnose and classify diabetic and non-diabetic patients. The main objective of this study was to provide an improved convolution neural network (CNN) model for automatic DR diagnosis from fundus images. The pooling function increases the receptive field of convolution kernels over layers. It reduces computational complexity and memory requirements because it reduces the resolution of feature maps while preserving the essential characteristics required for subsequent layer processing. In this study, an improved pooling function combined with an activation function in the ResNet-50 model was applied to the retina images in autonomous lesion detection with reduced loss and processing time. The improved ResNet-50 model was trained and tested over the two datasets (i.e., APTOS and Kaggle). The proposed model achieved an accuracy of 98.32% for APTOS and 98.71% for Kaggle datasets. It is proven that the proposed model has produced greater accuracy when compared to their state-of-the-art work in diagnosing DR with retinal fundus images.

Identification of tophi in ultrasound imaging based on transfer learning and clinical practice

Gout is a common metabolic disorder characterized by deposits of monosodium urate monohydrate crystals (tophi) in soft tissue, triggering intense and acute arthritis with intolerable pain as well as articular and periarticular inflammation. Tophi can also promote chronic inflammatory and erosive arthritis. 2015 ACR/EULAR Gout Classification criteria include clinical, laboratory, and imaging findings, where cases of gout are indicated by a threshold score of ≥ 8. Some imaging-related findings, such as a double contour sign in ultrasound, urate in dual-energy computed tomography, or radiographic gout-related erosion, generate a score of up to 4. Clearly, the diagnosis of gout is largely assisted by imaging findings; however, dual-energy computed tomography is expensive and exposes the patient to high levels of radiation. Although musculoskeletal ultrasound is non-invasive and inexpensive, the reliability of the results depends on expert experience. In the current study, we applied transfer learning to train a convolutional neural network for the identification of tophi in ultrasound images. The accuracy of predictions varied with the convolutional neural network model, as follows: InceptionV3 (0.871 ± 0.020), ResNet101 (0.913 ± 0.015), and VGG19 (0.918 ± 0.020). The sensitivity was as follows: InceptionV3 (0.507 ± 0.060), ResNet101 (0.680 ± 0.056), and VGG19 (0.747 ± 0.056). The precision was as follows: InceptionV3 (0.767 ± 0.091), ResNet101 (0.863 ± 0.098), and VGG19 (0.825 ± 0.062). Our results demonstrate that it is possible to retrain deep convolutional neural networks to identify the patterns of tophi in ultrasound images with a high degree of accuracy.

Artificial Learning and Machine Learning Applications in Spine Surgery: A Systematic Review

OBJECTIVES

This current systematic review sought to identify and evaluate all current research-based spine surgery applications of AI/ML in optimizing preoperative patient selection, as well as predicting and managing postoperative outcomes and complications.

METHODS

A comprehensive search of publications was conducted through the EMBASE, Medline, and PubMed databases using relevant keywords to maximize the sensitivity of the search. No limits were placed on level of evidence or timing of the study. Findings were reported according to the PRISMA guidelines.

RESULTS

After application of inclusion and exclusion criteria, 41 studies were included in this review. Bayesian networks had the highest average AUC (.80), and neural networks had the best accuracy (83.0%), sensitivity (81.5%), and specificity (71.8%). Preoperative planning/cost prediction models (.89,82.2%) and discharge/length of stay models (.80,78.0%) each reported significantly higher average AUC and accuracy compared to readmissions/reoperation prediction models (.67,70.2%) (P < .001, P = .005, respectively). Model performance also significantly varied across postoperative management applications for average AUC and accuracy values (P < .001, P < .027, respectively).

CONCLUSIONS

Generally, authors of the reviewed studies concluded that AI/ML offers a potentially beneficial tool for providers to optimize patient care and improve cost-efficiency. More specifically, AI/ML models performed best, on average, when optimizing preoperative patient selection and planning and predicting costs, hospital discharge, and length of stay. However, models were not as accurate in predicting postoperative complications, adverse events, and readmissions and reoperations. An understanding of AI/ML-based applications is becoming increasingly important, particularly in spine surgery, as the volume of reported literature, technology accessibility, and clinical applications continue to rapidly expand.

Opposition-Based Deer Hunting Optimization-Based Hybrid Classifier for Intrusion Detection in Wireless Sensor Networks

This paper tempts to implement a new machine-learning algorithm for detecting attacks in WSN. The developed model involves three main phases (a) Data Acquisition, (b) Feature Extraction, and (c) Detection. Next to the data acquisition from different benchmark datasets, the attributes in the form of features are extracted. Further, a new hybrid machine learning algorithm with the integration of Neural Network (NN), and Fuzzy Classifier is used for detection, and it is termed as FNN. As an improvement to the developed hybrid model, the number of hidden neurons in NN, and the membership function of Fuzzy Classifier is optimized by a modified optimization algorithm called Opposition-based Deer Hunting Optimization Algorithm (O-DHOA). Finally, the experiment analysis of our proposed model provides an effective solution to solve the problem of IDS detection and improves the performance of intrusion detection.

Enhanced Data Stream Classification by Optimized Weight Updated Meta-learning: Continuous learning-based on Concept-Drift

Purpose

The aim of continuous learning is to obtain and fine-tune information gradually without removing the already existing information. Many conventional approaches in streaming data classification assume that all arrived new data is completely labeled. To regularize Neural Networks (NNs) by merging side information like user-provided labels or pair-wise constraints, incremental semi-supervised learning models need to be introduced. However, they are hard to implement, specifically in non-stationary environments because of the efficiency and sensitivity of such algorithms to parameters. The periodic update and maintenance of the decision method is the significant challenge in incremental algorithms whenever the new data arrives.

Design/methodology/approach

Hence, this paper plans to develop the meta-learning model for handling continuous or streaming data. Initially, the data pertain to continuous behavior is gathered from diverse benchmark source. Further, the classification of the data is performed by the Recurrent Neural Network (RNN), in which testing weight is adjusted or optimized by the new meta-heuristic algorithm. Here, the weight is updated for reducing the error difference between the target and the measured data when new data is given for testing. The optimized weight updated testing is performed by evaluating the concept-drift and classification accuracy. The new continuous learning by RNN is accomplished by the improved Opposition-based Novel Updating Spotted Hyena Optimization (ONU-SHO). Finally, the experiments with different datasets show that the proposed learning is improved over the conventional models.

Findings

From the analysis, the accuracy of the ONU-SHO based RNN (ONU-SHO-RNN) was 10.1% advanced than Decision Tree (DT), 7.6% advanced than Naive Bayes (NB), 7.4% advanced than k-nearest neighbors (KNN), 2.5% advanced than Support Vector Machine (SVM) 9.3% advanced than NN, and 10.6% advanced than RNN. Hence, it is confirmed that the ONU-SHO algorithm is performing well for acquiring the best data stream classification.

Originality/value

This paper introduces a novel meta-learning model using Opposition-based Novel Updating Spotted Hyena Optimization (ONU-SHO)-based Recurrent Neural Network (RNN) for handling continuous or streaming data. This is the first work utilizes a novel meta-learning model using Opposition-based Novel Updating Spotted Hyena Optimization (ONU-SHO)-based Recurrent Neural Network (RNN) for handling continuous or streaming data.

Artificial intelligence in musculoskeletal ultrasound imaging

Ultrasonography (US) is noninvasive and offers real-time, low-cost, and portable imaging that facilitates the rapid and dynamic assessment of musculoskeletal components. Significant technological improvements have contributed to the increasing adoption of US for musculoskeletal assessments, as artificial intelligence (AI)-based computer-aided detection and computer-aided diagnosis are being utilized to improve the quality, efficiency, and cost of US imaging. This review provides an overview of classical machine learning techniques and modern deep learning approaches for musculoskeletal US, with a focus on the key categories of detection and diagnosis of musculoskeletal disorders, predictive analysis with classification and regression, and automated image segmentation. Moreover, we outline challenges and a range of opportunities for AI in musculoskeletal US practice.

Automated diagnosis of diabetic retinopathy enabled by optimized thresholding-based blood vessel segmentation and hybrid classifier

Objectives

The focus of this paper is to introduce an automated early Diabetic Retinopathy (DR) detection scheme from colour fundus images through enhanced segmentation and classification strategies by analyzing blood vessels.

Methods

The occurrence of DR is increasing from the past years, impacting the eyes due to a sudden rise in the glucose level of blood. All over the world, half of the people who are under age 70 are severely suffered from diabetes. The patients who are affected by DR will lose their vision during the absence of early recognition of DR and appropriate treatment. To decrease the growth and occurrence of loss of vision, the early detection and timely treatment of DR are desirable. At present, deep learning models have presented better performance using retinal images for DR detection. In this work, the input retinal fundus images are initially subjected to pre-processing that undergoes contrast enhancement by Contrast Limited Adaptive Histogram Equalization (CLAHE) and average filtering. Further, the optimized binary thresholding-based segmentation is done for blood vessel segmentation. For the segmented image, Tri-level Discrete Level Decomposition (Tri-DWT) is performed to decompose it. In the feature extraction phase, Local Binary Pattern (LBP), and Gray-Level Co-occurrence Matrices (GLCMs) are extracted. Next, the classification of images is done through the combination of two algorithms, one is Neural Network (NN), and the other Convolutional Neural Network (CNN). The extracted features are subjected to NN, and the tri-DWT-based segmented image is subjected to CNN. Both the segmentation and classification phases are enhanced by the improved meta-heuristic algorithm called Fitness Rate-based Crow Search Algorithm (FR-CSA), in which few parameters are optimized for attaining maximum detection accuracy.

Results

The proposed DR detection model was implemented in MATLAB 2018a, and the analysis was done using three datasets, HRF, Messidor, and DIARETDB.

Conclusions

The developed FR-CSA algorithm has the best detection accuracy in diagnosing DR.

Automated Maintenance Data Classification Using Recurrent Neural Network: Enhancement by Spotted Hyena-Based Whale Optimization

Data classification has been considered extensively in different fields, such as machine learning, artificial intelligence, pattern recognition, and data mining, and the expansion of classification has yielded immense achievements. The automatic classification of maintenance data has been investigated over the past few decades owing to its usefulness in construction and facility management. To utilize automated data classification in the maintenance field, a data classification model is implemented in this study based on the analysis of different mechanical maintenance data. The developed model involves four main steps: (a) data acquisition, (b) feature extraction, (c) feature selection, and (d) classification. During data acquisition, four types of dataset are collected from the benchmark Google datasets. The attributes of each dataset are further processed for classification. Principal component analysis and first-order and second-order statistical features are computed during the feature extraction process. To reduce the dimensions of the features for error-free classification, feature selection was performed. The hybridization of two algorithms, the Whale Optimization Algorithm (WOA) and Spotted Hyena Optimization (SHO), tends to produce a new algorithm—i.e., a Spotted Hyena-based Whale Optimization Algorithm (SH-WOA), which is adopted for performing feature selection. The selected features are subjected to a deep learning algorithm called Recurrent Neural Network (RNN). To enhance the efficiency of conventional RNNs, the number of hidden neurons in an RNN is optimized using the developed SH-WOA. Finally, the efficacy of the proposed model is verified utilizing the entire dataset. Experimental results show that the developed model can effectively solve uncertain data classification, which minimizes the execution time and enhances efficiency.

Modeling and Analysis of Ultrasound Elastographic Axial Strains for Spine Fracture Identification

This study reports the first use of ultrasound (US) elastography for imaging spinal fractures by assessing the mechanical response of the soft tissue at the posterior vertebra boundary to a uniaxial compression in rabbit ex vivo samples. Three-dimensional finite-element (FE) models of the vertebra-soft tissue complex in rabbit samples are generated and analyzed to evaluate the distribution of the axial normal and shear strains at the vertebra-soft tissue interface. Experiments on the same samples are performed to corroborate simulation findings. Results of this study indicate that the distribution of the axial strains manifests as distinct patterns around intact and fractured vertebrae. Numerical characteristics of the axial strain's spatial distribution are further used to construct two shape descriptors to make inferences on spinal abnormalities: 1) axial normal strain asymmetry for assessing the presence of fractures and 2) principal orientation of axial shear strain concentration regions (shear zones) for measurement of spinous process dislocation. This study demonstrates that axial normal strain and axial shear strain maps obtained via US elastography can provide a new means to detect spine fractures and abnormalities in the selected ex vivo animal models. Spinal fracture detection is important for the assessment of spinal cord injuries and stability. However, identification of spinal fractures using US is currently challenging. Our results show that features resulting from strain elastograms can serve as a useful adjunct to B-mode images in identifying spine fractures in the selected animal samples, and this information could be helpful in clinical settings.

Enabling pregnant women and their physicians to make informed medication decisions using artificial intelligence

The role of artificial intelligence (AI) in healthcare for pregnant women. To assess the role of AI in women’s health, discover gaps, and discuss the future of AI in maternal health. A systematic review of English articles using EMBASE, PubMed, and SCOPUS. Search terms included pregnancy and AI. Research articles and book chapters were included, while conference papers, editorials and notes were excluded from the review. Included papers focused on pregnancy and AI methods, and pertained to pharmacologic interventions. We identified 376 distinct studies from our queries. A final set of 31 papers were included for the review. Included papers represented a variety of pregnancy concerns and multidisciplinary applications of AI. Few studies relate to pregnancy, AI, and pharmacologics and therefore, we review carefully those studies. External validation of models and techniques described in the studies is limited, impeding on generalizability of the studies. Our review describes how AI has been applied to address maternal health, throughout the pregnancy process: preconception, prenatal, perinatal, and postnatal health concerns. However, there is a lack of research applying AI methods to understand how pharmacologic treatments affect pregnancy. We identify three areas where AI methods could be used to improve our understanding of pharmacological effects of pregnancy, including: (a) obtaining sound and reliable data from clinical records (15 studies), (b) designing optimized animal experiments to validate specific hypotheses (1 study) to (c) implementing decision support systems that inform decision-making (11 studies). The largest literature gap that we identified is with regards to using AI methods to optimize translational studies between animals and humans for pregnancy-related drug exposures.

A novel approach to neuraxial anesthesia: application of an automated ultrasound spinal landmark identification

Background

Neuraxial procedures are commonly performed for therapeutic and diagnostic indications. Currently, they are typically performed via palpation-guided surface landmark. We devised a novel intelligent image processing system that identifies spinal landmarks using ultrasound images. Our primary aim was to evaluate the first attempt success rate of spinal anesthesia using landmarks obtained from the automated spinal landmark identification technique.

Methods

In this prospective cohort study, we recruited 100 patients who required spinal anesthesia for surgical procedures. The video from ultrasound scan image of the L3/4 interspinous space in the longitudinal view and the posterior complex in the transverse view were recorded. The demographic and clinical characteristics were collected and analyzed based on the success rates of the spinal insertion.

Results

Success rate (95%CI) for dural puncture at first attempt was 92.0% (85.0–95.9%). Median time to detection of posterior complex was 45.0 [IQR: 21.9, 77.3] secs. There is good correlation observed between the program-recorded depth and the clinician-measured depth to the posterior complex (r = 0.94).

Conclusions

The high success rate and short time taken to obtain the surface landmark with this novel automated ultrasound guided technique could be useful to clinicians to utilise ultrasound guided neuraxial techniques with confidence to identify the anatomical landmarks on the ultrasound scans. Future research would be to define the use in more complex patients during the administration of neuraxial blocks.

Trial registration

This study was retrospectively registered on clinicaltrials.gov registry (NCT03535155) on 24 May 2018.

Machine Learning in Orthopedics: A Literature Review

In this paper we present the findings of a systematic literature review covering the articles published in the last two decades in which the authors described the application of a machine learning technique and method to an orthopedic problem or purpose. By searching both in the Scopus and Medline databases, we retrieved, screened and analyzed the content of 70 journal articles, and coded these resources following an iterative method within a Grounded Theory approach. We report the survey findings by outlining the articles' content in terms of the main machine learning techniques mentioned therein, the orthopedic application domains, the source data and the quality of their predictive performance.

Machine learning for medical ultrasound: status, methods, and future opportunities

Ultrasound (US) imaging is the most commonly performed cross-sectional diagnostic imaging modality in the practice of medicine. It is low-cost, non-ionizing, portable, and capable of real-time image acquisition and display. US is a rapidly evolving technology with significant challenges and opportunities. Challenges include high inter- and intra-operator variability and limited image quality control. Tremendous opportunities have arisen in the last decade as a result of exponential growth in available computational power coupled with progressive miniaturization of US devices. As US devices become smaller, enhanced computational capability can contribute significantly to decreasing variability through advanced image processing. In this paper, we review leading machine learning (ML) approaches and research directions in US, with an emphasis on recent ML advances. We also present our outlook on future opportunities for ML techniques to further improve clinical workflow and US-based disease diagnosis and characterization.

Automatic Localization of the Needle Target for Ultrasound-Guided Epidural Injections

Accurate identification of the needle target is crucial for effective epidural anesthesia. Currently, epidural needle placement is administered by a manual technique, relying on the sense of feel, which has a significant failure rate. Moreover, misleading the needle may lead to inadequate anesthesia, post dural puncture headaches, and other potential complications. Ultrasound offers guidance to the physician for identification of the needle target, but accurate interpretation and localization remain challenges. A hybrid machine learning system is proposed to automatically localize the needle target for epidural needle placement in ultrasound images of the spine. In particular, a deep network architecture along with a feature augmentation technique is proposed for automatic identification of the anatomical landmarks of the epidural space in ultrasound images. Experimental results of the target localization on planes of 3-D as well as 2-D images have been compared against an expert sonographer. When compared with the expert annotations, the average lateral and vertical errors on the planes of 3-D test data were 1 and 0.4 mm, respectively. On 2-D test data set, an average lateral error of 1.7 mm and vertical error of 0.8 mm were acquired.

Development of a real-time lumbar ultrasound image processing system for epidural needle entry site localization

A fully-automatic ultrasound image processing system that can determine the needle entry site for epidural anesthesia (EA) in real time is presented in this paper. Neither the knowledge of anesthetists nor additional hardware is required to operate the system, which firstly directs the anesthetists to reach the desired insertion region in the longitudinal view, i.e., lumbar level L3-L4, and then locates the ideal puncture site by instructing the anesthetists to rotate and slightly adjust the position of ultrasound probe. In order to implement these functions, modules including image processing, panorama stitching, feature extraction/selection, template matching and support vector machine (SVM) classification are incorporated in this system. Additionally, a user-friendly graphical user interface (GUI), which displays the processing results and guides anesthetists intuitively, is further designed to conceal the intricacy of algorithms. Feasibility and effectiveness of the proposed system has been evaluated through a set of realtime tests on 53 volunteers from a local hospital.

SLIDE: automatic spine level identification system using a deep convolutional neural network

PURPOSE

Percutaneous spinal needle insertion procedures often require proper identification of the vertebral level to effectively and safely deliver analgesic agents. The current clinical method involves "blind" identification of the vertebral level through manual palpation of the spine, which has only 30% reported accuracy. Therefore, there is a need for better anatomical identification prior to needle insertion.

METHODS

A real-time system was developed to identify the vertebral level from a sequence of ultrasound images, following a clinical imaging protocol. The system uses a deep convolutional neural network (CNN) to classify transverse images of the lower spine. Several existing CNN architectures were implemented, utilizing transfer learning, and compared for adequacy in a real-time system. In the system, the CNN output is processed, using a novel state machine, to automatically identify vertebral levels as the transducer moves up the spine. Additionally, a graphical display was developed and integrated within 3D Slicer. Finally, an augmented reality display, projecting the level onto the patient's back, was also designed. A small feasibility study [Formula: see text] evaluated performance.

RESULTS

The proposed CNN successfully discriminates ultrasound images of the sacrum, intervertebral gaps, and vertebral bones, achieving 88% 20-fold cross-validation accuracy. Seventeen of 20 test ultrasound scans had successful identification of all vertebral levels, processed at real-time speed (40 frames/s).

CONCLUSION

A machine learning system is presented that successfully identifies lumbar vertebral levels. The small study on human subjects demonstrated real-time performance. A projection-based augmented reality display was used to show the vertebral level directly on the subject adjacent to the puncture site.