Neurogenic and Myogenic Diseases: Quantitative Texture Analysis of Muscle US Data for Differentiation

Machine learning-based radiomics to differentiate immune-mediated necrotizing myopathy from limb-girdle muscular dystrophy R2 using MRI

Objectives

This study aimed to assess the feasibility of a machine learning-based radiomics tools to discriminate between Limb-girdle muscular dystrophy R2 (LGMDR2) and immune-mediated necrotizing myopathy (IMNM) using lower-limb muscle magnetic resonance imaging (MRI) examination.

Methods

After institutional review board approval, 30 patients with genetically proven LGMDR2 (12 females; age, 34.0 ± 11.3) and 45 patients with IMNM (28 females; age, 49.2 ± 16.6) who underwent lower-limb MRI examination including T1-weighted and interactive decomposition water and fat with echos asymmetric and least-squares estimation (IDEAL) sequences between July 2014 and August 2022 were included. Radiomics features of muscles were obtained, and four machine learning algorithms were conducted to select the optimal radiomics classifier for differential diagnosis. This selected algorithm was performed to construct the T1-weighted (TM), water-only (WM), or the combined model (CM) for calf-only, thigh-only, or the calf and thigh MR images, respectively. And their diagnostic performance was studied using area under the curve (AUC) and compared to the semi-quantitative model constructed by the modified Mercuri scale of calf and thigh muscles scored by two radiologists specialized in musculoskeletal imaging.

Results

The logistic regression (LR) model was the optimal radiomics model. The performance of the WM and CM for thigh-only images (AUC 0.893, 0.913) was better than those for calf-only images (AUC 0.846, 0.880) except the TM. For "calf + thigh" images, the TM, WM, and CM models always performed best (AUC 0.953, 0.907, 0.953) with excellent accuracy (92.0, 84.0, 88.0%). The AUCs of the Mercuri model of the calf, thigh, and "calf + thigh" images were 0.847, 0.900, and 0.953 with accuracy (84.0, 84.0, 88.0%).

Conclusion

Machine learning-based radiomics models can differentiate LGMDR2 from IMNM, performing better than visual assessment. The model built by combining calf and thigh images presents excellent diagnostic efficiency.

Assessment and risk prediction of frailty using texture-based muscle ultrasound image analysis and machine learning techniques

The purpose of this study was to evaluate texture-based muscle ultrasound image analysis for the assessment and risk prediction of frailty phenotype. This retrospective study of prospectively acquired data included 101 participants who underwent ultrasound scanning of the anterior thigh. Participants were subdivided according to frailty phenotype and were followed up for two years. Primary and secondary outcome measures were death and comorbidity, respectively. Forty-three texture features were computed from the rectus femoris and the vastus intermedius muscles using statistical methods. Model performance was evaluated by computing the area under the receiver operating characteristic curve (AUC) while outcome prediction was evaluated using regression analysis. Models developed achieved a moderate to good AUC (0.67 ≤ AUC ≤ 0.79) for categorizing frailty. The stepwise multiple logistic regression analysis demonstrated that they correctly classified 70-87% of the cases. The models were associated with increased comorbidity (0.01 ≤ p ≤ 0.18) and were predictive of death for pre-frail and frail participants (0.001 ≤ p ≤ 0.016). In conclusion, texture analysis can be useful to identify frailty and assess risk prediction (i.e. mortality) using texture features extracted from muscle ultrasound images in combination with a machine learning approach.

Feasibility of artificial intelligence assisted quantitative muscle ultrasound in carpal tunnel syndrome

BACKGROUND

In case of focal neuropathy, the muscle fibers innervated by the corresponding nerves are replaced with fat or fibrous tissue due to denervation, which results in increased echo intensity (EI) on ultrasonography. EI analysis can be conducted quantitatively using gray scale analysis. Mean value of pixel brightness of muscle image defined as EI. However, the accuracy achieved by using this parameter alone to differentiate between normal and abnormal muscles is limited. Recently, attempts have been made to increase the accuracy using artificial intelligence (AI) in the analysis of muscle ultrasound images. CTS is the most common disease among focal neuropathy. In this study, we aimed to verify the utility of AI assisted quantitative analysis of muscle ultrasound in CTS.

METHODS

This is retrospective study that used data from adult who underwent ultrasonographic examination of hand muscles. The patient with CTS confirmed by electromyography and subjects without CTS were included. Ultrasound images of the unaffected hands of patients or subjects without CTS were used as controls. Ultrasonography was performed by one physician in same sonographic settings. Both conventional quantitative grayscale analysis and machine learning (ML) analysis were performed for comparison.

RESULTS

A total of 47 hands with CTS and 27 control hands were analyzed. On conventional quantitative analysis, mean EI ratio (i.e. mean thenar EI/mean hypothenar EI ratio) were significantly higher in the patient group than in the control group, and the AUC was 0.76 in ROC analysis. In the analysis using machine learning, the AUC was the highest for the linear support vector classifier (AUC = 0.86). When recursive feature elimination was applied to the classifier, the AUC value improved to 0.89.

CONCLUSION

This study showed a significant increase in diagnostic accuracy when AI was used for quantitative analysis of muscle ultrasonography. If an analysis protocol using machine learning can be established and mounted on an ultrasound machine, a noninvasive and non-time-consuming muscle ultrasound examination can be conducted as an ancillary tool for diagnosis.

Muscle magnetic resonance characterization of STIM1 tubular aggregate myopathy using unsupervised learning

PURPOSE

Congenital myopathies are a heterogeneous group of diseases affecting the skeletal muscles and characterized by high clinical, genetic, and histological variability. Magnetic Resonance (MR) is a valuable tool for the assessment of involved muscles (i.e., fatty replacement and oedema) and disease progression. Machine Learning is becoming increasingly applied for diagnostic purposes, but to our knowledge, Self-Organizing Maps (SOMs) have never been used for the identification of the patterns in these diseases. The aim of this study is to evaluate if SOMs may discriminate between muscles with fatty replacement (S), oedema (E) or neither (N).

METHODS

MR studies of a family affected by tubular aggregates myopathy (TAM) with the histologically proven autosomal dominant mutation of the STIM1 gene, were examined: for each patient, in two MR assessments (i.e., t0 and t1, the latter after 5 years), fifty-three muscles were evaluated for muscular fatty replacement on the T1w images, and for oedema on the STIR images, for reference. Sixty radiomic features were collected from each muscle at t0 and t1 MR assessment using 3DSlicer software, in order to obtain data from images. A SOM was created to analyze all datasets using three clusters (i.e., 0, 1 and 2) and results were compared with radiological evaluation.

RESULTS

Six patients with TAM STIM1-mutation were included. At t0 MR assessments, all patients showed widespread fatty replacement that intensifies at t1, while oedema mainly affected the muscles of the legs and appears stable at follow-up. All muscles with oedema showed fatty replacement, too. At t0 SOM grid clustering shows almost all N muscles in Cluster 0 and most of the E muscles in Cluster 1; at t1 almost all E muscles appear in Cluster 1.

CONCLUSION

Our unsupervised learning model appears to be able to recognize muscles altered by the presence of edema and fatty replacement.

Hypothesis generation: Quantitative research to levator ani muscle injury based on MRI texture analysis

AIM

Patients with pelvic organ prolapse (POP) mostly have injury to the levator ani muscle (LAM). We aimed to assess LAM injury in POP patients by quantifying texture feature (TF) ratios between the LAM and the obturator internus muscle (OIM) using texture analysis.

METHODS

This study retrospectively enrolled 32 participants, including 24 patients with POP and eight people with normal pelvic floor muscles. TFs of the LAM and the OIM were extracted using LIFEx version 6.30, and an independent samples t-test was performed to determine TF ratios characterizing LAM injury. After dimension reduction and binary logic analysis, the optimal TF ratio was obtained and the LAM injury quantitative evaluation was proposed. Spearman's correlation was performed to explore the correlations between TF ratios and clinical characteristics. We compared the diagnostic performance of quantitative evaluation and visual evaluation.

RESULTS

There were significant differences in 13 TF ratios between the POP and control groups. The area under the receiver operating characteristic curve of the integrated TF ratio was 0.948. Integrated TF ratio was significantly correlated with body mass index, pregnancies, and vaginal deliveries but had no correlation with LAM volume, hiatal area or abortions. Compared with the visual evaluation, the diagnostic accuracy of the quantitative evaluation had improved by 63.2% and 14.3% in the "minor defect" and "major defect" categories, respectively.

CONCLUSION

The integrated TF ratio can be used as a new quantifiable index to characterize LAM injury. The TF evaluation provides a potential role in LAM injury noninvasive diagnostic.

The diversity of hereditary neuromuscular diseases: Experiences from molecular diagnosis

BACKGROUND

Hereditary neuromuscular diseases (NMDs) are a group of rare disorders, and the diagnosis of these diseases is a substantial burden for referral centers. Although next-generation sequencing (NGS) has identified a large number of genes associated with hereditary NMDs, the diagnostic rates still vary across centers.

METHODS

Patients with a suspected hereditary NMD were referred to neuromuscular specialists at the National Taiwan University Hospital. Molecular diagnoses were performed by employing a capture panel containing 194 genes associated with NMDs.

RESULTS

Among the 50 patients referred, 43 had a suspicion of myopathy, and seven had polyneuropathy. The overall diagnostic rate was 58%. Pathogenic variants in 19 genes were observed; the most frequent pathogenic variant found in this cohort (DYSF) was observed in only four patients, and 10 pathogenic variants were observed in one patient each. One case of motor neuron disease was clinically mistaken for myopathy. A positive family history increased the diagnostic rate (positive: 72.7% vs. negative: 56.3%). Fourteen patients with elevated plasma creatine kinase levels remained without a diagnosis.

CONCLUSION

The application of NGS in this single-center study proves the great diversity of hereditary NMDs. A capture panel is essential, but high-quality clinical and laboratory evaluations of patients are also indispensable.

Evaluation of lateral pterygoid muscle in patients with temporomandibular joint anterior disk displacement using T1-weighted Dixon sequence: a retrospective study

Background

Pathological alterations of lateral pterygoid muscle (LPM) are implicated in temporomandibular joint anterior disk displacement (ADD). However, quantification of the fatty infiltration of LPM and its correlation with ADD have rarely been reported. The aim of this study was to evaluate the fatty infiltration, morphological features and texture features of LPM in patients with ADD using T1-weighted Dixon sequence.

Methods

This retrospective study included patients who underwent temporomandibular joint MRI with T1-weighted Dixon sequence between December 2018 and August 2020. The temporomandibular joints of the included patients were divided into three groups according to the position of disk: Normal position disk (NP) group, Anterior disk displacement with reduction (ADDWR) group and Anterior disk displacement without reduction (ADDWOR) group. Fat fraction, morphological features (Length; Width; Thickness), and texture features (Angular second moment; Contrast; Correlation; Inverse different moment; Entropy) extracted from in-phase image of LPM were evaluated. One-way ANOVA, Welch’s ANOVA, Kruskal–Wallis test, Spearman and Pearson correlation analysis were performed. Intra-class correlation coefficient was used to evaluate the reproducibility.

Results

A total of 53 patients with 106 temporomandibular joints were evaluated. Anterior disk displacement without reduction group showed higher fat fraction than normal position disk group (P = 0.024). Length of LPM was negatively correlated with fat fraction (r = -0.22, P = 0.026). Angular second moment (ρ = -0.32, P < 0.001), correlation (ρ = -0.28, P = 0.003) and inverse different moment (ρ = -0.27, P = 0.005) were negatively correlated with fat fraction, while positive correlation was found between entropy and fat fraction (ρ = 0.31, P = 0.001). The intra-class correlation coefficients for all values were ranged from 0.80 to 0.97.

Conclusions

Patients with ADDWOR present more fatty infiltration in the LPM compared to NP or ADDWR patients. Fatty infiltration of LPM was associated with more atrophic and higher intramuscular heterogeneity in patients with ADD. Fat fraction of LPM quantitatively and noninvasively evaluated by Dixon sequence may has utility as an imaging-based marker of the structural severity of ADD disease process, which could be clinical helpful for the early diagnose of ADD and predication of disease progression.

The Application of Texture Feature Analysis of Rectus Femoris Based on Local Binary Pattern (LBP) Combined With Gray-Level Co-Occurrence Matrix (GLCM) in Sarcopenia

OBJECTIVES

In order to detect the changes in muscle texture of sarcopenia and to explore a new method of ultrasound assessment of muscle changes in sarcopenia.

METHODS

we used the local binary pattern (LBP) combined with gray-level co-occurrence matrix (GLCM) method to extract and quantitatively analyze the texture information of the rectus femoris of different people, and initially verified the robustness of this method to image gain changes. We recruited young volunteers, elderly volunteers without sarcopenia, and elderly volunteers with sarcopenia in this cross-sectional study. We scanned the rectus femoris and extracted their muscle texture features.

RESULTS

We found that when ultrasonographic gain varied from 40% to 70%, the intraclass correlation coefficient (ICC) of contrast, entropy, and homogeneity were 0.989, 0.973, and 0.989, respectively. Body mass index was significantly related to contrast (r = 0.285, P < .05), and age had a significant correlation with contrast and homogeneity (r = -0.259 and r = 0.269, P < .05). The elderly volunteers with sarcopenia had the highest entropy (0.363 [0.342-0.403]) and homogeneity (2.203 [2.162-2.277]) in the texture of the rectus femoris among the three groups, and at the same time had the lowest contrast (44.583 [43.492-47.399]), and all P < .05.

CONCLUSION

LBP combined with GLCM can be a stable method for extracting muscle texture features. At the same time, the contrast, entropy, and homogeneity of the rectus femoris of the elderly with sarcopenia were significantly different from those of the young volunteers and the elderly without sarcopenia, suggesting the texture features of rectus femoris are potential parameters for evaluating muscle function and pathological changes.

Diagnostic Performance of Muscle Echo Intensity and Fractal Dimension for the Detection of Frailty Phenotype

To determine the relationship between muscle echo intensity (EI) and fractal dimension (FD), and the diagnostic performance of both ultrasound parameters for the identification of frailty phenotype. A retrospective interpretation of ultrasound scans from a previous cohort (November 2014-February 2015) was performed. The sample included healthy participants <60 years old, and participants ≥60 divided into robust, pre-frail, and frail groups according to Fried frailty criteria. A region of interest of the rectus femoris from the ultrasound scan was segmented, and histogram function was applied to obtain EI. For fractal analysis, images were processed using two-dimensional box-counting techniques to calculate FD. Statistical analyses were performed with diagnostic performance tests. A total of 102 participants (mean age 63 ± 16, 57 men) were evaluated. Muscle fractal dimension correlated with EI (r = .38, p < .01) and showed different pattern in the scatter plots when participants were grouped by non-frail (control + robust) and frail (pre-frail + frail). The diagnostic accuracy for EI to categorize frailty was of 0.69 (95%CI: 0.59-0.78, p = .001), with high intra-rater (ICC: 0.98, 95%CI: 0.98-0.99); p < .001) and inter-rater (ICC: 0.89, 95%CI: 0.75-0.95; p < .001) reliability and low measurement error for both parameters (EI: -0.18, LOA95%: -10.8 to 10.5; FD: 0.00, LOA95%: -0.09 to 0.10) in arbitrary units. The ROC curve combining both parameters was not better than EI alone (p = .18). Muscle FD correlated with EI and showed different patterns according to frailty phenotype, with EI outperforming FD as a possible diagnostic tool for frailty.

MRI and muscle imaging for idiopathic inflammatory myopathies

Although idiopathic inflammatory myopathies (IIM) are a heterogeneous group of diseases nearly all patients display muscle inflammation. Originally, muscle biopsy was considered as the gold standard for IIM diagnosis. The development of muscle imaging led to revisiting not only the IIM diagnosis strategy but also the patients' follow-up. Different techniques have been tested or are in development for IIM including positron emission tomography, ultrasound imaging, ultrasound shear wave elastography, though magnetic resonance imaging (MRI) remains the most widely used technique in routine. Whereas guidelines on muscle imaging in myositis are lacking here we reviewed the relevance of muscle imaging for both diagnosis and myositis patients' follow-up. We propose recommendations about when and how to perform MRI on myositis patients, and we describe new techniques that are under development.

Texture analysis of muscle MRI: machine learning-based classifications in idiopathic inflammatory myopathies

To develop a machine learning (ML) model that predicts disease groups or autoantibodies in patients with idiopathic inflammatory myopathies (IIMs) using muscle MRI radiomics features. Twenty-two patients with dermatomyositis (DM), 14 with amyopathic dermatomyositis (ADM), 19 with polymyositis (PM) and 19 with non-IIM were enrolled. Using 2D manual segmentation, 93 original features as well as 93 local binary pattern (LBP) features were extracted from MRI (short-tau inversion recovery [STIR] imaging) of proximal limb muscles. To construct and compare ML models that predict disease groups using each set of features, dimensional reductions were performed using a reproducibility analysis by inter-reader and intra-reader correlation coefficients, collinearity analysis, and the sequential feature selection (SFS) algorithm. Models were created using the linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), support vector machine (SVM), k-nearest neighbors (k-NN), random forest (RF) and multi-layer perceptron (MLP) classifiers, and validated using tenfold cross-validation repeated 100 times. We also investigated whether it was possible to construct models predicting autoantibody status. Our ML-based MRI radiomics models showed the potential to distinguish between PM, DM, and ADM. Models using LBP features provided better results, with macro-average AUC values of 0.767 and 0.714, accuracy of 61.2 and 61.4%, and macro-average recall of 61.9 and 59.8%, in the LDA and k-NN classifiers, respectively. In contrast, the accuracies of radiomics models distinguishing between non-IIM and IIM disease groups were low. A subgroup analysis showed that classification models for anti-Jo-1 and anti-ARS antibodies provided AUC values of 0.646–0.853 and 0.692–0.792, with accuracy of 71.5–81.0 and 65.8–78.3%, respectively. ML-based TA of muscle MRI may be used to predict disease groups or the autoantibody status in patients with IIM and is useful in non-invasive assessments of disease mechanisms.

Muscle ultrasound: Present state and future opportunities

Muscle ultrasound is a valuable addition to the neuromuscular toolkit in both the clinic and research settings, with proven value and reliability. However, it is currently not fulfilling its full potential in the diagnostic care of patients with neuromuscular disease. This review highlights the possibilities and pitfalls of muscle ultrasound as a diagnostic tool and biomarker, and discusses challenges to its widespread implementation. We expect that limitations in visual image interpretation, posed by user inexperience, could be overcome with simpler scoring systems and the help of deep-learning algorithms. In addition, more information should be collected on the relation between specific neuromuscular disorders, disease stages, and expected ultrasound abnormalities, as this will enhance specificity of the technique and enable the use of muscle ultrasound as a biomarker. Quantified muscle ultrasound gives the most sensitive results but is hampered by the need for device-specific reference values. Efforts in creating dedicated muscle ultrasound systems and artificial intelligence to help with image interpretation are expected to improve usability. Finally, the standard inclusion of muscle and nerve ultrasound in neuromuscular teaching curricula and guidelines will facilitate further implementation in practice. Our hope is that this review will help in unleashing muscle ultrasound's full potential.

Muscle Composition Analysis of Ultrasound Images: A Narrative Review of Texture Analysis

Skeletal muscle composition, often characterized by the degree of intramuscular adipose tissue, deteriorates with aging and disease and contributes to impairments in function and metabolism. Ultrasound can provide surrogate measures of muscle composition through measurement of echo intensity; however, there are several limitations associated with its analysis. More complex image processing features, broadly known as texture analysis, can also provide surrogates of muscle composition and may circumvent some of the limitations associated with muscle echo intensity. Here, texture features from the intensity histogram, gray-level co-occurrence matrix, run-length matrix, local binary pattern, blob analysis, texture anisotropy index and wavelet analysis are discussed. The purpose of this review was to provide a conceptual understanding of texture analysis as it pertains to muscle composition of ultrasound images.

Association of Thigh Muscle Strength with Texture Features Based on Proton Density Fat Fraction Maps Derived from Chemical Shift Encoding-Based Water-Fat MRI

Purpose: Based on conventional and quantitative magnetic resonance imaging (MRI), texture analysis (TA) has shown encouraging results as a biomarker for tissue structure. Chemical shift encoding-based water–fat MRI (CSE-MRI)-derived proton density fat fraction (PDFF) of thigh muscles has been associated with musculoskeletal, metabolic, and neuromuscular disorders and was demonstrated to predict muscle strength. The purpose of this study was to investigate PDFF-based TA of thigh muscles as a predictor of thigh muscle strength in comparison to mean PDFF. Methods: 30 healthy subjects (age = 30 ± 6 years; 15 females) underwent CSE-MRI of the lumbar spine at 3T, using a six-echo 3D spoiled gradient echo sequence. Quadriceps (EXT) and ischiocrural (FLEX) muscles were segmented to extract mean PDFF and texture features. Muscle flexion and extension strength were measured with an isokinetic dynamometer. Results: Of the eleven extracted texture features, Variance(global) showed the highest significant correlation with extension strength (p < 0.001, R²_adj = 0.712), and Correlation showed the highest significant correlation with flexion strength (p = 0.016, R²_adj = 0.658). Multivariate linear regression models identified Variance(global) and sex, but not PDFF, as significant predictors of extension strength (R²_adj = 0.709; p < 0.001), while mean PDFF, sex, and BMI, but none of the texture features, were identified as significant predictors of flexion strength (R²_adj = 0.674; p < 0.001). Conclusions: Prediction of quadriceps muscle strength can be improved beyond mean PDFF by means of TA, indicating the capability to quantify muscular fat infiltration patterns.

Texture Features of Proton Density Fat Fraction Maps from Chemical Shift Encoding-Based MRI Predict Paraspinal Muscle Strength

Texture analysis (TA) has shown promise as a surrogate marker for tissue structure, based on conventional and quantitative MRI sequences. Chemical-shift-encoding-based MRI (CSE-MRI)-derived proton density fat fraction (PDFF) of paraspinal muscles has been associated with various medical conditions including lumbar back pain (LBP) and neuromuscular diseases (NMD). Its application has been shown to improve the prediction of paraspinal muscle strength beyond muscle volume. Since mean PDFF values do not fully reflect muscle tissue structure, the purpose of our study was to investigate PDFF-based TA of paraspinal muscles as a predictor of muscle strength, as compared to mean PDFF. We performed 3T-MRI of the lumbar spine in 26 healthy subjects (age = 30 ± 6 years; 15 females) using a six-echo 3D spoiled gradient echo sequence for chemical-shift-encoding-based water–fat separation. Erector spinae (ES) and psoas (PS) muscles were segmented bilaterally from level L2–L5 to extract mean PDFF and texture features. Muscle flexion and extension strength was measured with an isokinetic dynamometer. Out of the eleven texture features extracted for each muscle, Kurtosis(global) of ES showed the highest significant correlation (r = 0.59, p = 0.001) with extension strength and Variance(global) of PS showed the highest significant correlation (r = 0.63, p = 0.001) with flexion strength. Using multivariate linear regression models, Kurtosis(global) of ES and BMI were identified as significant predictors of extension strength (R²_adj = 0.42; p < 0.001), and Variance(global) and Skewness(global) of PS were identified as significant predictors of flexion strength (R²_adj = 0.59; p = 0.001), while mean PDFF was not identified as a significant predictor. TA of CSE-MRI-based PDFF maps improves the prediction of paraspinal muscle strength beyond mean PDFF, potentially reflecting the ability to quantify the pattern of muscular fat infiltration. In the future, this may help to improve the pathophysiological understanding, diagnosis, monitoring and treatment evaluation of diseases with paraspinal muscle involvement, e.g., NMD and LBP.

Texture Analysis on Ultrasound: The Effect of Time Gain Compensation on Histogram Metrics and Gray-Level Matrices

Aims:

Texture analysis (TA) is becoming an increasingly used tool in radiological research. Some papers have been published on its use in ultrasound (US), but the way in which the machine settings affect the features has not yet been fully explored. With this research, we analyze how the time gain compensation (TGC) influences the features of the gray-level matrices in the abdominal US setting.

Subjects and Methods:

We analyzed the images acquired from the hepatorenal acoustic window of a healthy 29-year-old volunteer acquired with different TGC settings. TA was carried out using the LifeX software.

Results:

Several both 1^st and 2^nd order gray-level matrices features showed a strong correlation with TGC settings.

Conclusions:

TGC settings must be accounted for when carrying out further TA studies.

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.

Quantitative Ultrasound Texture Analysis to Assess the Spastic Muscles in Stroke Patients

This study aimed to investigate the feasibility of sonoelastography for determining echotexture in post-stroke patients. Moreover, the relationships of muscle echotexture features, muscle stiffness, and functional performance in spastic muscle were explored. The study population comprised 22 males with stroke. The echotexture features (entropy and energy) of the biceps brachii muscles (BBM) in both arms were extracted by local binary pattern (LBP) from ultrasound images, whereas the stiffness of BBM was assessed by shear wave velocity (SWV) in the transverse and longitudinal planes. The Fugl–Meyer assessment (FMA) was used to assess the functional performance of the upper arm. The results showed that echotexture was more inhomogeneous in the paretic BBM than in the non-paretic BBM. SWV was significantly faster in paretic BBM than in non-paretic BBM. Both echotexture features were significantly correlated with SWV in the longitudinal plane. The feature of energy was significantly negatively correlated with FMA in the longitudinal plane and was significantly positively correlated with the duration from stroke onset in the transverse plane. The echotexture extracted by LBP may be a promising approach for quantitative assessment of the spastic BBM in post-stroke patients.

Musculoskeletal ultrasound for traumatic and torsional alterations

The sonographic appearance of soft tissue can be altered by trauma and positional change with torsional stress. This creates challenges for ultrasonographic interpretation, because most descriptive literature and standard instructional references are displayed in anatomically neutral or other conventional positions. Knowledge of anatomic alteration and changes in sonographic appearance with torsional stress is essential for accurately assessing soft tissue abnormalities in conditions of spasticity, traumatic and post-surgical changes, and other conditions that distort musculoskeletal relationships. A systematic scanning approach to these alterations is needed for accurate diagnostic interpretation, optimizing electrode placement for electrodiagnostic techniques, effective needle placement for therapeutic ultrasound-guided procedures, and even planning for restorative surgery. This review describes expected positional changes of normal structures with torsional alteration, as well as sonographic recognition of scars, burns, hematomas, fat layer fracture, Morel-Lavallee lesions, abscesses, foreign bodies, myotendinous lesions, muscle injury and denervation, and traumatic peripheral nerve injury.

Texture analysis of sonographic muscle images can distinguish myopathic conditions

Given the recent technological advent of muscle ultrasound (US), classification of various myopathic conditions could be possible, especially by mathematical analysis of muscular fine structure called texture analysis. We prospectively enrolled patients with three neuromuscular conditions and their lower leg US images were quantitatively analyzed by texture analysis and machine learning methodology in the following subjects :  Inclusion body myositis (IBM) [N=11] ; myotonic dystrophy type 1 (DM1) [N=19] ; polymyositis/dermatomyositis (PM-DM) [N=21]. Although three-group analysis achieved up to 58.8% accuracy, two-group analysis of IBM plus PM-DM versus DM1 showed 78.4% accuracy. Despite the small number of subjects, texture analysis of muscle US followed by machine learning might be expected to be useful in identifying myopathic conditions. J. Med. Invest. 66 : 237-240, August, 2019.

A focused review of myokines as a potential contributor to muscle hypertrophy from resistance-based exercise

PURPOSE

Resistance exercise induces muscle growth and is an important treatment for age-related losses in muscle mass and strength. Myokines are hypothesized as a signal conveying physiological information to skeletal muscle, possibly to "fine-tune" other regulatory pathways. While myokines are released from skeletal muscle following contraction, their role in increasing muscle mass and strength in response to resistance exercise or training is not established. Recent research identified both local and systemic release of myokines after an acute bout of resistance exercise. However, it is not known whether myokines with putative anabolic function are mechanistically involved in producing muscle hypertrophy after resistance exercise. Further, nitric oxide (NO), an important mediator of muscle stem cell activation, upregulates the expression of certain myokine genes in skeletal muscle.

METHOD

In the systemic context of complex hypertrophic signaling, this review: (1) summarizes literature on several well-recognized, representative myokines with anabolic potential; (2) explores the potential mechanistic role of myokines in skeletal muscle hypertrophy; and (3) identifies future research required to advance our understanding of myokine anabolism specifically in skeletal muscle.

RESULT

This review establishes a link between myokines and NO production, and emphasizes the importance of considering systemic release of potential anabolic myokines during resistance exercise as complementary to other signals that promote hypertrophy.

CONCLUSION

Investigating adaptations to resistance exercise in aging opens a novel avenue of interdisciplinary research into myokines and NO metabolites during resistance exercise, with the longer-term goal to improve muscle health in daily living, aging, and rehabilitation.

Quantitative CT texture analysis for diagnosing systemic sclerosis: Effect of iterative reconstructions and radiation doses

To test whether texture analysis (TA) can discriminate between Systemic Sclerosis (SSc) and non-SSc patients in computed tomography (CT) with different radiation doses and reconstruction algorithms.In this IRB-approved retrospective study, 85 CT scans at different radiation doses [49 standard dose CT (SDCT) with a volume CT dose index (CTDIvol) of 4.86 ± 2.1 mGy and 36 low-dose (LDCT) with a CTDIvol of 2.5 ± 1.5 mGy] were selected; 61 patients had Ssc ("cases"), and 24 patients had no SSc ("controls"). CT scans were reconstructed with filtered-back projection (FBP) and with sinogram-affirmed iterative reconstruction (SAFIRE) algorithms. 304 TA features were extracted from each manually drawn region-of-interest at 6 pre-defined levels: at the midpoint between lung apices and tracheal carina, at the level of the tracheal carina, and 4 between the carina and pleural recesses. Each TA feature was averaged between these 6 pre-defined levels and was used as input in the machine learning algorithm artificial neural network (ANN) with backpropagation (MultilayerPerceptron) for differentiating between SSc and non-SSc patients.Results were compared regarding correctly/incorrectly classified instances and ROC-AUCs.ANN correctly classified individuals in 93.8% (AUC = 0.981) of FBP-LDCT, in 78.5% (AUC = 0.859) of FBP-SDCT, in 91.1% (AUC = 0.922) of SAFIRE3-LDCT and 75.7% (AUC = 0.815) of SAFIRE3-SDCT, in 88.1% (AUC = 0.929) of SAFIRE5-LDCT and 74% (AUC = 0.815) of SAFIRE5-SDCT.Quantitative TA-based discrimination of CT of SSc patients is possible showing highest discriminatory power in FBP-LDCT images.

Texture Analysis and Machine Learning for Detecting Myocardial Infarction in Noncontrast Low-Dose Computed Tomography: Unveiling the Invisible

OBJECTIVES

The aim of this study was to test whether texture analysis and machine learning enable the detection of myocardial infarction (MI) on non-contrast-enhanced low radiation dose cardiac computed tomography (CCT) images.

MATERIALS AND METHODS

In this institutional review board-approved retrospective study, we included non-contrast-enhanced electrocardiography-gated low radiation dose CCT image data (effective dose, 0.5 mSv) acquired for the purpose of calcium scoring of 27 patients with acute MI (9 female patients; mean age, 60 ± 12 years), 30 patients with chronic MI (8 female patients; mean age, 68 ± 13 years), and in 30 subjects (9 female patients; mean age, 44 ± 6 years) without cardiac abnormality, hereafter termed controls. Texture analysis of the left ventricle was performed using free-hand regions of interest, and texture features were classified twice (Model I: controls versus acute MI versus chronic MI; Model II: controls versus acute and chronic MI). For both classifications, 6 commonly used machine learning classifiers were used: decision tree C4.5 (J48), k-nearest neighbors, locally weighted learning, RandomForest, sequential minimal optimization, and an artificial neural network employing deep learning. In addition, 2 blinded, independent readers visually assessed noncontrast CCT images for the presence or absence of MI.

RESULTS

In Model I, best classification results were obtained using the k-nearest neighbors classifier (sensitivity, 69%; specificity, 85%; false-positive rate, 0.15). In Model II, the best classification results were found with the locally weighted learning classification (sensitivity, 86%; specificity, 81%; false-positive rate, 0.19) with an area under the curve from receiver operating characteristics analysis of 0.78. In comparison, both readers were not able to identify MI in any of the noncontrast, low radiation dose CCT images.

CONCLUSIONS

This study indicates the ability of texture analysis and machine learning in detecting MI on noncontrast low radiation dose CCT images being not visible for the radiologists' eye.

Advances in Quantitative Imaging of Genetic and Acquired Myopathies: Clinical Applications and Perspectives

In the last years, magnetic resonance imaging (MRI) has become fundamental for the diagnosis and monitoring of myopathies given its ability to show the severity and distribution of pathology, to identify specific patterns of damage distribution and to properly interpret a number of genetic variants. The advances in MR techniques and post-processing software solutions have greatly expanded the potential to assess pathological changes in muscle diseases, and more specifically of myopathies; a number of features can be studied and quantified, ranging from composition, architecture, mechanical properties, perfusion, and function, leading to what is known as quantitative MRI (qMRI). Such techniques can effectively provide a variety of information beyond what can be seen and assessed by conventional MR imaging; their development and application in clinical practice can play an important role in the diagnostic process and in assessing disease course and treatment response. In this review, we briefly discuss the current role of muscle MRI in diagnosing muscle diseases and describe in detail the potential and perspectives of the application of advanced qMRI techniques in this field.

Predicting Survival Using Pretreatment CT for Patients With Hepatocellular Carcinoma Treated With Transarterial Chemoembolization: Comparison of Models Using Radiomics

OBJECTIVE

The purpose of this study was to investigate the use of radiomics features as prognostic biomarkers for predicting the survival of patients treated with transarterial chemoembolization (TACE) for hepatocellular carcinoma (HCC).

MATERIALS AND METHODS

We retrospectively analyzed 88 patients with HCC treated with TACE. High-dimensional quantitative feature analysis was applied to extract 116 radiomics features of pretreatment CT. A radiomics score model was constructed from these features with the use of least absolute shrinkage and selection operator Cox regression. A clinical score model was constructed from clinical variables with the use of multivariate Cox regression. A combined score model was constructed using the radiomics and clinical models. We compared the three models (the radiomics score, clinical score, and combined score models) for predicting overall survival, using Kaplan-Meier analysis and the log-rank test.

RESULTS

The following radiomics features were selected for the radiomics score model: histogram-based features (median, kurtosis, and energy), shape-based features (spherical disproportion and surface-to-volume ratio), gray-level co-occurrence matrix (GLCM)-based features (energy, informational measure of correlation, maximum probability, contrast, and sum average), and intensity size zone matrix-based features (size zone variability). For the clinical score model, the Child-Pugh score, α-fetoprotein level, and HCC size were included. The combined score model included five radiomics features (surface area-to-volume ratio, kurtosis, median, gray-level co-occurrence matrix contrast, and size zone variability) and three clinical factors (Child-Pugh score, α-fetoprotein level, and HCC size). The combined model was a better predictor of survival (hazard ratio, 19.88; p < 0.0001) than the clinical score model or the radiomics score model.

CONCLUSION

A radiomics approach combined with conventional clinical variables could be effective in predicting the survival of patients with HCC treated with TACE.

Texture analysis of paraspinal musculature in MRI of the lumbar spine: analysis of the lumbar stenosis outcome study (LSOS) data

OBJECTIVE

To evaluate association of fatty infiltration in paraspinal musculature with clinical outcomes in patients suffering from lumbar spinal stenosis (LSS) using qualitative and quantitative grading in magnetic resonance imaging (MRI).

MATERIALS AND METHODS

In this retrospective study, texture analysis (TA) was performed on postprocessed axial T2 weighted (w) MR images at level L3/4 using dedicated software (MaZda) in 62 patients with LSS. Associations in fatty infiltration between qualitative Goutallier and quantitative TA findings with two clinical outcome measures, Spinal stenosis measure (SSM) score and walking distance, at baseline and regarding change over time were assessed using machine learning algorithms and multiple logistic regression models.

RESULTS

Quantitative assessment of fatty infiltration using the histogram TA feature "mean" showed higher interreader reliability (ICC 0.83-0.97) compared to the Goutallier staging (κ = 0.69-0.93). No correlation between Goutallier staging and clinical outcome measures was observed. Among 151 TA features, only TA feature "mean" of the spinotransverse group showed a significant but weak correlation with worsened SSM (p = 0.046). TA feature "S(3,3) entropy" showed a significant but weak association with worsened WD over 12 months (p = 0.046).

CONCLUSION

MR TA is a reproducible tool to quantitatively assess paraspinal fatty infiltration, but there is no clear association with the clinical outcome in asymptomatic LSS patients.

Prediction of successful shock wave lithotripsy with CT: a phantom study using texture analysis

OBJECTIVE

To apply texture analysis (TA) in computed tomography (CT) of urinary stones and to correlate TA findings with the number of required shockwaves for successful shock wave lithotripsy (SWL).

MATERIALS AND METHODS

CT was performed on thirty-four urinary stones in an in vitro setting. Urinary stones underwent SWL and the number of required shockwaves for disintegration was recorded. TA was performed after post-processing for pixel spacing and image normalization. Feature selection and dimension reduction were performed according to inter- and intrareader reproducibility and by evaluating the predictive ability of the number of shock waves with the degree of redundancy between TA features. Three regression models were tested: (1) linear regression with elimination of colinear attributes (2), sequential minimal optimization regression (SMOreg) employing machine learning, and (3) simple linear regression model of a single TA feature with lowest squared error.

RESULTS

Highest correlations with the absolute number of required SWL shockwaves were found for the linear regression model (r = 0.55, p = 0.005) using two weighted TA features: Histogram 10th Percentile, and Gray-Level Co-Occurrence Matrix (GLCM) S(3, 3) SumAverg. Using the median number of required shockwaves (n = 72) as a threshold, receiver-operating characteristic analysis showed largest area-under-the-curve values for the SMOreg model (AUC = 0.84, r = 0.51, p < 0.001) using four weighted TA features: Histogram 10^th Percentile, and GLCM S(1, 1) InvDfMom, S(3, 3) SumAverg, and S(4, -4) SumVarnc.

CONCLUSION

Our in vitro study illustrates the proof-of-principle of TA of urinary stone CT images for predicting the success of stone disintegration with SWL.

Local Texture Anisotropy as an Estimate of Muscle Quality in Ultrasound Imaging

This study introduces local pattern texture anisotropy as a novel parameter to differentiate healthy and disordered muscle and to gauge the severity of muscle impairments based on B-mode ultrasound images. Preliminary human results are also presented. A local pattern texture anisotropy index (TAI) was computed in one region of interest in the short head of the biceps brachii. The effects of gain settings and box sizes required for TAI computation were investigated. Between-day reliability was studied in patients with sporadic inclusion body myositis (n = 26). The ability of the TAI to discriminate dystrophic from healthy muscle was evaluated in patients with Duchenne muscular dystrophy and healthy controls (n = 16). TAI values were compared with a gray-scale index (GSI). TAI values were less influenced by gain settings than were GSI values. TAI had lower between-day variability (typical error = 2.3%) compared with GSI (typical error = 2.3% vs. 8.3%, respectively). Patients with Duchenne muscular dystrophy had lower TAIs than controls (0.76 ± 0.06 vs. 0.87 ± 0.03, respectively, p <0.05). At 40% gain, TAI values correlated with percentage predicted elbow flexor strength in inclusion body myositis (R = 0.63, p <0.001). The TAI may be a promising addition to other texture-based approaches for quantitative muscle ultrasound imaging.

Histogram analysis derived from apparent diffusion coefficient (ADC) is more sensitive to reflect serological parameters in myositis than conventional ADC analysis

OBJECTIVE

Diffusion-weighted imaging (DWI) has the potential of being able to reflect histopathology architecture. A novel imaging approach, namely histogram analysis, is used to further characterize tissues on MRI. The aim of this study was to correlate histogram parameters derived from apparent diffusion coefficient (ADC) maps with serological parameters in myositis.

METHODS

16 patients with autoimmune myositis were included in this retrospective study. DWI was obtained on a 1.5 T scanner by using the b-values of 0 and 1000 s mm^-2. Histogram analysis was performed as a whole muscle measurement by using a custom-made Matlab-based application. The following ADC histogram parameters were estimated: ADCmean, ADCmax, ADCmin, ADCmedian, ADCmode, and the following percentiles ADCp10, ADCp25, ADCp75, ADCp90, as well histogram parameters kurtosis, skewness, and entropy. In all patients, the blood sample was acquired within 3 days to the MRI. The following serological parameters were estimated: alanine aminotransferase, aspartate aminotransferase, creatine kinase, lactate dehydrogenase, C-reactive protein (CRP) and myoglobin. All patients were screened for Jo1-autobodies.

RESULTS

Kurtosis correlated inversely with CRP (p = -0.55 and 0.03). Furthermore, ADCp10 and ADCp90 values tended to correlate with creatine kinase (p = -0.43, 0.11, and p = -0.42, = 0.12 respectively). In addition, ADCmean, p10, p25, median, mode, and entropy were different between Jo1-positive and Jo1-negative patients.

CONCLUSION

ADC histogram parameters are sensitive for detection of muscle alterations in myositis patients. Advances in knowledge: This study identified that kurtosis derived from ADC maps is associated with CRP in myositis patients. Furthermore, several ADC histogram parameters are statistically different between Jo1-positive and Jo1-negative patients.

Automated diagnosis of myositis from muscle ultrasound: Exploring the use of machine learning and deep learning methods

OBJECTIVE

To evaluate the use of ultrasound coupled with machine learning (ML) and deep learning (DL) techniques for automated or semi-automated classification of myositis.

METHODS

Eighty subjects comprised of 19 with inclusion body myositis (IBM), 14 with polymyositis (PM), 14 with dermatomyositis (DM), and 33 normal (N) subjects were included in this study, where 3214 muscle ultrasound images of 7 muscles (observed bilaterally) were acquired. We considered three problems of classification including (A) normal vs. affected (DM, PM, IBM); (B) normal vs. IBM patients; and (C) IBM vs. other types of myositis (DM or PM). We studied the use of an automated DL method using deep convolutional neural networks (DL-DCNNs) for diagnostic classification and compared it with a semi-automated conventional ML method based on random forests (ML-RF) and "engineered" features. We used the known clinical diagnosis as the gold standard for evaluating performance of muscle classification.

RESULTS

The performance of the DL-DCNN method resulted in accuracies ± standard deviation of 76.2% ± 3.1% for problem (A), 86.6% ± 2.4% for (B) and 74.8% ± 3.9% for (C), while the ML-RF method led to accuracies of 72.3% ± 3.3% for problem (A), 84.3% ± 2.3% for (B) and 68.9% ± 2.5% for (C).

CONCLUSIONS

This study demonstrates the application of machine learning methods for automatically or semi-automatically classifying inflammatory muscle disease using muscle ultrasound. Compared to the conventional random forest machine learning method used here, which has the drawback of requiring manual delineation of muscle/fat boundaries, DCNN-based classification by and large improved the accuracies in all classification problems while providing a fully automated approach to classification.

The Potential Value of Preoperative MRI Texture and Shape Analysis in Grading Meningiomas: A Preliminary Investigation

OBJECT: Preoperative knowledge of meningioma grade is essential for planning treatment and surgery. The purpose of this study was to investigate the diagnostic value of MRI texture and shape analysis in grading meningiomas. METHODS: A surgical database was reviewed to identify meningioma patients who had undergone tumor resection between January 2015 and December 2016. Preoperative MR images were retrieved and analyzed. Texture and shape analysis was conducted to quantitatively evaluate tumor heterogeneity and morphology. Three machine learning classifiers were trained with these features to build classification models. The performance of the features and classification models was assessed. RESULTS: A total of 131 patients were included in this study: 21 with high-grade meningiomas and 110 with low-grade meningiomas. Three texture features were selected: Horzl_RLNonUni, S(2,2)SumOfSqs, and WavEnHL_s-3; three shape features were selected: GeoFv, GeoW4, and GeoW5b. The Mann–Whitney test indicated that all six features were significantly different between high-grade and low-grade meningiomas. AUC values were generally greater than 0.50 (range, 0.73 to 0.88). Sensitivities and specificities ranged from 47.62% to 90.48% and 69.09% to 96.36%, respectively. Among the nine classification models obtained, the one built by training the SVM classifier with all six features achieved the best performance, with a sensitivity, specificity, diagnostic accuracy, and AUC of 0.86, 0.87, 0.87, and 0.87, respectively. CONCLUSIONS: Texture and shape analysis, especially when combined with a SVM classifier, can provide satisfactory performance in the preoperative determination of meningioma grade and is thus potentially useful for clinical application.