Residual one-dimensional convolutional neural network for neuromuscular disorder classification from needle electromyography signals with explainability

BACKGROUND AND OBJECTIVE

Neuromuscular disorders are diseases that damage our ability to control body movements. Needle electromyography (nEMG) is often used to diagnose neuromuscular disorders, which is an electrophysiological test measuring electric signals generated from a muscle using an invasive needle. Characteristics of nEMG signals are manually analyzed by an electromyographer to diagnose the types of neuromuscular disorders, and this process is highly dependent on the subjective experience of the electromyographer. Contemporary computer-aided methods utilized deep learning image classification models to classify nEMG signals which are not optimized for classifying signals. Additionally, model explainability was not addressed which is crucial in medical applications. This study aims to improve prediction accuracy, inference time, and explain model predictions in nEMG neuromuscular disorder classification.

METHODS

This study introduces the nEMGNet, a one-dimensional convolutional neural network with residual connections designed to extract features from raw signals with higher accuracy and faster speed compared to image classification models from previous works. Next, the divide-and-vote (DiVote) algorithm was designed to integrate each subject's heterogeneous nEMG signal data structures and to utilize muscle subtype information for higher accuracy. Finally, feature visualization was used to identify the causality of nEMGNet diagnosis predictions, to ensure that nEMGNet made predictions on valid features, not artifacts.

RESULTS

The proposed method was tested using 376 nEMG signals measured from 57 subjects between June 2015 to July 2020 in Seoul National University Hospital. The results from the three-class classification task demonstrated that nEMGNet's prediction accuracy of nEMG signal segments was 62.35%, and the subject diagnosis prediction accuracy of nEMGNet and the DiVote algorithm was 83.69 %, over 5-fold cross-validation. nEMGNet outperformed all models from previous works on nEMG diagnosis classification, and heuristic analysis of feature visualization results indicate that nEMGNet learned relevant nEMG signal characteristics.

CONCLUSIONS

This study introduced nEMGNet and DiVote algorithm which demonstrated fast and accurate performance in predicting neuromuscular disorders based on nEMG signals. The proposed method may be applied in medicine to support real-time electrophysiologic diagnosis.

Preoperative evaluation of mammographic microcalcifications after neoadjuvant chemotherapy for breast cancer

AIM

To assess the predictive value of preoperative residual mammographic microcalcifications for residual tumours after neoadjuvant chemotherapy (NAC) for breast cancer.

MATERIALS AND METHODS

This single-centre retrospective study included breast cancer patients who underwent NAC and demonstrated suspicious microcalcifications within or near the tumour bed on mammography from June 2015 to August 2018. The residual microcalcifications and remnant lesion on magnetic resonance imaging (MRI) were correlated with histopathological findings of residual tumours and immunohistochemical markers.

RESULTS

A total of 96 patients were included. Ten patients achieved pathological complete response (pCR) and previous suspicious microcalcifications were associated with benign pathology in 10.4% (10/96) of the patients. In the remaining 86 patients who did not achieve pCR, 61.5% (59/96) of the residual microcalcifications were associated with invasive or in situ carcinoma and 28.1% (27/96) with benign pathology. Hormone receptor-positive (HR⁺) patients had the highest proportion of residual malignant microcalcifications compared to HR^- patients (48.9% versus 13.5%, respectively; p=0.019). MRI correlated better than residual microcalcifications on mammography in predicting residual tumour extent in all subtypes (ICC=0.709 versus 0.365). MRI also showed higher correlation with residual tumour size for the HR^-/HER2⁺ and HR^-/HER2^- subtype (ICC=0.925 and 0.876, respectively).

CONCLUSION

The extent of microcalcifications on mammography after NAC did not correlate with the extent of residual cancer in 38.5% of women. Regardless of the extent of microcalcifications, residual tumour extent on MRI after NAC and molecular subtype could be an accurate tool in evaluating residual cancer after NAC.

Oxygen, nitric oxide and articular cartilage

Molecular oxygen is required for the production of nitric oxide (NO), a pro-inflammatory mediator that is associated with osteoarthritis and rheumatoid arthritis. To date there has been little consideration of the role of oxygen tension in the regulation of nitric oxide production associated with arthritis. Oxygen tension may be particularly relevant to articular cartilage since it is avascular and therefore exists at a reduced oxygen tension. The superficial zone exists at approximately 6% O2, while the deep zone exists at less than 1% O2. Furthermore, oxygen tension can alter matrix synthesis, and the material properties of articular cartilage in vitro. The increase in nitric oxide associated with arthritis can be caused by pro-inflammatory cytokines and mechanical stress. Oxygen tension significantly alters endogenous NO production in articular cartilage, as well as the stimulation of NO in response to both mechanical loading and pro-inflammatory cytokines. Mechanical loading and pro-inflammatory cytokines also increase the production of prostaglandin E2 (PGE2). There is a complex interaction between NO and PGE2, and oxygen tension can alter this interaction. These findings suggest that the relatively low levels of oxygen within the joint may have significant influences on the metabolic activity, and inflammatory response of cartilage as compared to ambient levels. A better understanding of the role of oxygen in the production of inflammatory mediators in response to mechanical loading, or pro-inflammatory cytokines, may aid in the development of strategies for therapeutic intervention in arthritis.

Pathogenesis of osteoarthritis-like changes in the joints of mice deficient in type IX collagen

OBJECTIVE

To examine the pathogenetic mechanisms of osteoarthritis (OA)-like changes in Col9a1-/- mice, which are deficient in type IX collagen.

METHODS

Knee joints and temporomandibular joints (TMJs) from Col9a1-/- mice and their wild-type (Col9a1+/+) littermates were examined by light microscopy. Immunohistochemical staining was performed to examine the expression of matrix metalloproteinase 3 (MMP-3) and MMP-13, degraded type II collagen, and the discoidin domain receptor 2 (DDR-2) in knee joints. Cartilage mechanics were also evaluated for compressive properties by microindentation testing of the tibial plateau and for tensile properties by osmotic loading of the femoral condyle.

RESULTS

Histologic analysis showed age-dependent OA-like changes in the knee and TMJs of Col9a1-/- mice starting at the age of 3 months. At the age of 6 months, enhanced proteoglycan degradation was observed in the articular cartilage of the knee and TMJs of the mutant mice. The expression of MMP-13 and DDR-2 protein and the amount of degraded type II collagen were higher in the knee joints of Col9a1-/- mice than in their wild-type littermates at the age of 6 months. Changes in cartilage mechanics were observed in the femoral and tibial plateaus of Col9a1-/- mice at 6 months, including a decrease in the compressive modulus and uniaxial modulus. At 3 and 6 months of age, tibial cartilage in Col9a1-/- mice was found to be more permeable to fluid flow, with an associated compromise in the fluid pressurization mechanism of load support. All of these changes occurred only at medial sites.

CONCLUSION

Lack of type IX collagen in Col9a1-/- mice results in age-dependent OA-like changes in the knee joints and TMJs.

Zonal variations in the three-dimensional morphology of the chondron measured in situ using confocal microscopy

OBJECTIVE

Chondrocytes in articular cartilage are surrounded by a narrow pericellular matrix (PCM), which together with the enclosed cell(s) are termed the "chondron". Although the precise function of this tissue region is unknown, previous studies provide indirect evidence that the PCM plays an important role in governing the local mechanical environment of chondrocytes. In particular, theoretical models of the chondron under mechanical loading suggest that the shape, size, and biomechanical properties of the PCM significantly influence the stress-strain and fluid flow environment of the cell. The goal of this study was to quantify the three-dimensional morphology of chondron in situ using en bloc immunolabeling of type VI collagen coupled with fluorescence confocal microscopy.

METHODS

Three-dimensional reconstructions of intact, fluorescently labeled chondrons were made from stacks of confocal images recorded in situ from the superficial, middle, and deep zones of porcine articular cartilage of the medial femoral condyle.

RESULTS

Significant variations in the shape, size, and orientation of chondrocytes and chondrons were observed with depth from the tissue surface, revealing flattened discoidal chondrons in the superficial zone, rounded chondrons in the middle zone, and elongated, multicellular chondrons in the deep zone.

CONCLUSIONS

The shape and orientation of the chondron appear to reflect the local collagen architecture of the interterritorial matrix, which varies significantly with depth. Quantitative measurements of morphology of the chondron and its variation with site, disease, or aging may provide new insights into the influence of this structure on physiology and the pathology of articular cartilage.

An in situ calibration of an ultrasound transducer: a potential application for an ultrasonic indentation test of articular cartilage

A change in mechanical properties of articular cartilage would be considered one of the most reliable signs of cartilage degeneration. While an indentation method has the potential to measure the cartilage properties in vivo, an accurate measurement of cartilage thickness in situ is technically difficult. An ultrasound transducer has often been used to measure the cartilage thickness. However, its accuracy is limited by the lack of an accurate measurement of the ultrasound speed of cartilage, for the ultrasound speed varies according to the pathological conditions of the tissue. Therefore, the objective of this study is to develop an in situ calibration method of predicting the true ultrasound speed of cartilage and thus allow the ultrasound transducer to measure the thickness of the tissue with great accuracy. By simultaneously implementing an indentation testing protocol using the ultrasound transducer as an indenter, this method can also provide an indentation stiffness measurement of cartilage. The feasibility of the proposed method was examined using normal and proteoglycan-depleted cartilage specimens. It was found that the true ultrasound speed measured by the in situ calibration method was sensitive to the proteoglycan depletion (1735+/-35 m/s for normal, and 1598+/-28 m/s for proteoglycan-depleted cartilage), and that the measured cartilage thickness was consistently accurate regardless of the tissue condition. The measured indentation stiffness of articular cartilage was also sensitive to the tissue condition. Thus, this study demonstrates that the proposed ultrasonic indentation technique can be used to accurately identify the abnormality of articular cartilage in situ.