A computer-aided diagnosis of multiple sclerosis based on mfVEP recordings

Multimodal diagnostics in multiple sclerosis: predicting disability and conversion from relapsing-remitting to secondary progressive disease course - protocol for systematic review and meta-analysis

BACKGROUND

The number of patients diagnosed with multiple sclerosis (MS) has increased significantly over the last decade. The challenge is to identify the transition from relapsing-remitting to secondary progressive MS. Since available methods to examine patients with MS are limited, both the diagnostics and prognostication of disease progression would benefit from the multimodal approach. The latter combines the evidence obtained from disparate radiologic modalities, neurophysiological evaluation, cognitive assessment and molecular diagnostics. In this systematic review we will analyse the advantages of multimodal studies in predicting the risk of conversion to secondary progressive MS.

METHODS AND ANALYSIS

We will use peer-reviewed publications available in Web of Science, Medline/PubMed, Scopus, Embase and CINAHL databases. In vivo studies reporting the predictive value of diagnostic methods will be considered. Selected publications will be processed through Covidence software for automatic deduplication and blind screening. Two reviewers will use a predefined template to extract the data from eligible studies. We will analyse the performance metrics (1) for the classification models reflecting the risk of secondary progression: sensitivity, specificity, accuracy, area under the receiver operating characteristic curve, positive and negative predictive values; (2) for the regression models forecasting disability scores: the ratio of mean absolute error to the range of values. Then, we will create ranking charts representing performance of the algorithms for calculating disability level and MS progression. Finally, we will compare the predictive power of radiological and radiomical correlates of clinical disability and cognitive impairment in patients with MS.

ETHICS AND DISSEMINATION

The study does not require ethical approval because we will analyse publicly available literature. The project results will be published in a peer-review journal and presented at scientific conferences.

PROSPERO REGISTRATION NUMBER

CRD42022354179.

Multiple Sclerosis Diagnosis Using Machine Learning and Deep Learning: Challenges and Opportunities

Multiple Sclerosis (MS) is a disease that impacts the central nervous system (CNS), which can lead to brain, spinal cord, and optic nerve problems. A total of 2.8 million are estimated to suffer from MS. Globally, a new case of MS is reported every five minutes. In this review, we discuss the proposed approaches to diagnosing MS using machine learning (ML) published between 2011 and 2022. Numerous models have been developed using different types of data, including magnetic resonance imaging (MRI) and clinical data. We identified the methods that achieved the best results in diagnosing MS. The most implemented approaches are SVM, RF, and CNN. Moreover, we discussed the challenges and opportunities in MS diagnosis to improve AI systems to enable researchers and practitioners to enhance their approaches and improve the automated diagnosis of MS. The challenges faced by automated MS diagnosis include difficulty distinguishing the disease from other diseases showing similar symptoms, protecting the confidentiality of the patients' data, achieving reliable ML models that are also easily understood by non-experts, and the difficulty of collecting a large reliable dataset. Moreover, we discussed several opportunities in the field such as the implementation of secure platforms, employing better AI solutions, developing better disease prognosis systems, combining more than one data type for better MS prediction and using OCT data for diagnosis, utilizing larger, multi-center datasets to improve the reliability of the developed models, and commercialization.

A Hybrid Approach for MS Diagnosis Through Nonlinear EEG Descriptors and Metaheuristic Optimized Classification Learning

Multiple sclerosis (MS), a disease of the central nervous system, affects the white matter of the brain. Neurologists interpret magnetic resonance images that are often complicated, time-consuming, and contradictory. Using EEG signals, this disease can be analyzed and diagnosed more accurately. However, it is imperative that MS be diagnosed by specialists using assistive technology. Until now, a few methods have been proposed in this field that are sometimes associated with different challenges in analysis. This paper presents a hybrid approach to MS diagnosis in order to decrease classification error rates. Using the proposed method, EEG descriptors in both the time and frequency domains are analyzed. After the feature extraction stage, a modified version of the ant colony optimization method (m-ACO) was used to select the appropriate subset of features. Then, the support vector machine is used to determine whether the disease exists. A metaheuristic algorithm adjusts the support vector machine's parameters to withstand overfitting challenges. Despite a limited number of input channels, significant classification accuracy has been achieved using wavelet analysis techniques, dividing all five subbands of EEG signals, signal windowing, and extracting efficient features from the data. Additionally, alpha, beta, and gamma bands of the signal are important, and the accuracy, sensitivity, and specificity levels are higher than 98.5%. Compared to similar MS diagnostic methods, the proposed method achieved significantly higher diagnostic accuracy. Application and implementation of this method can be effective in treating neurological diseases, including multiple sclerosis.

Current review and next steps for artificial intelligence in multiple sclerosis risk research

In the last few decades, the prevalence of multiple sclerosis (MS), a chronic inflammatory disease of the nervous system, has increased, particularly in Northern European countries, the United States, and United Kingdom. The promise of artificial intelligence (AI) and machine learning (ML) as tools to address problems in MS research has attracted increasing interest in these methods. Bayesian networks offer a clear advantage since they can integrate data and causal knowledge allowing for visualizing interactions between dependent variables and potential confounding factors. A review of AI/ML research methods applied to MS found 216 papers using terms "Multiple Sclerosis", "machine learning", "artificial intelligence", "Bayes", and "Bayesian", of which 90 were relevant and recently published. More than half of these involve the detection and segmentation of MS lesions for quantitative analysis; however clinical and lifestyle risk factor assessment and prediction have largely been ignored. Of those that address risk factors, most provide only association studies for some factors and often fail to include the potential impact of confounding factors and bias (especially where these have causal explanations) that could affect data interpretation, such as reporting quality and medical care access in various countries. To address these gaps in the literature, we propose a causal Bayesian network approach to assessing risk factors for MS, which can address deficiencies in current epidemiological methods of producing risk measurements and makes better use of observational data.

Using advanced analysis of multifocal visual-evoked potentials to evaluate the risk of clinical progression in patients with radiologically isolated syndrome

This study aimed to assess the role of multifocal visual-evoked potentials (mfVEPs) as a guiding factor for clinical conversion of radiologically isolated syndrome (RIS). We longitudinally followed a cohort of 15 patients diagnosed with RIS. All subjects underwent thorough ophthalmological, neurological and imaging examinations. The mfVEP signals were analysed to obtain features in the time domain (SNR_min: amplitude, Lat_max: monocular latency) and in the continuous wavelet transform (CWT) domain (b_max: instant in which the CWT function maximum appears, N_max: number of CWT function maximums). The best features were used as inputs to a RUSBoost boosting-based sampling algorithm to improve the mfVEP diagnostic performance. Five of the 15 patients developed an objective clinical symptom consistent with an inflammatory demyelinating central nervous system syndrome during follow-up (mean time: 13.40 months). The (SNR_min) variable decreased significantly in the group that converted (2.74 ± 0.92 vs. 4.07 ± 0.95, p = 0.01). Similarly, the (b_max) feature increased significantly in RIS patients who converted (169.44 ± 24.81 vs. 139.03 ± 11.95 (ms), p = 0.02). The area under the curve analysis produced SNR_min and b_max values of 0.92 and 0.88, respectively. These results provide a set of new mfVEP features that can be potentially useful for predicting prognosis in RIS patients.

Identification of clusters in multifocal electrophysiology recordings to maximize discriminant capacity (patients vs. control subjects)

PURPOSE

To propose a new method of identifying clusters in multifocal electrophysiology (multifocal electroretinogram: mfERG; multifocal visual-evoked potential: mfVEP) that conserve the maximum capacity to discriminate between patients and control subjects.

METHODS

The theoretical framework proposed creates arbitrary N-size clusters of sectors. The capacity to discriminate between patients and control subjects is assessed by analysing the area under the receiver operator characteristic curve (AUC). As proof of concept, the method is validated using mfERG recordings taken from both eyes of control subjects (n = 6) and from patients with multiple sclerosis (n = 15).

RESULTS

Considering the amplitude of wave P1 as the analysis parameter, the maximum value of AUC = 0.7042 is obtained with N = 9 sectors. Taking into account the AUC of the amplitudes and latencies of waves N1 and P1, the maximum value of the AUC = 0.6917 with N = 8 clustered sectors. The greatest discriminant capacity is obtained by analysing the latency of wave P1: AUC = 0.8854 with a cluster of N = 12 sectors.

CONCLUSION

This paper demonstrates the effectiveness of a method able to determine the arbitrary clustering of multifocal responses that possesses the greatest capacity to discriminate between control subjects and patients when applied to the visual field of mfERG or mfVEP recordings. The method may prove helpful in diagnosing any disease that is identifiable in patients' mfERG or mfVEP recordings and is extensible to other clinical tests, such as optical coherence tomography.