Machine Learning Techniques for Differential Diagnosis of Vertigo and Dizziness: A Review

Visual scale to document acute dizziness in the hospital

BACKGROUND

Managing acute vertigo/dizziness for inpatients requires valid communication between the various healthcare professionals that triage such life-threatening presentations, yet there are no current scaling methods for managing such acute vertigo symptoms for inpatients.

OBJECTIVE

To describe the development and validation of the Krems Acute Vertigo/Dizziness Scale (KAVEDIS), a new instrument for tracking subjective symptoms (vertigo, dizziness) and gait impairment across four unique vestibular diagnoses (Menière's disease, benign paroxysmal positional vertigo, peripheral vestibular hypofunction, and vestibular migraine) over a one-year period after inpatient hospital admission.

METHODS

Retrospective data collection study from KAVEDIS scale and chart documentation.

RESULTS

The KAVEDIS scale can significantly distinguish scores from admission to discharge in three of four vestibular diagnoses. The documented course of subjective vestibular symptoms and gait disturbances were correlated in all four groups.

CONCLUSION

We suggest that KAVEDIS documentation among inpatients admitted with acute vertigo/dizziness may improve communication between the various intervening clinicians and help to raise concern in cases of symptomprogression.

Machine learning models help differentiate between causes of recurrent spontaneous vertigo

BACKGROUND

Vestibular migraine (VM) and Menière's disease (MD) are two common causes of recurrent spontaneous vertigo. Using history, video-nystagmography and audiovestibular tests, we developed machine learning models to separate these two disorders.

METHODS

We recruited patients with VM or MD from a neurology outpatient facility. One hundred features from six "feature subsets": history, acute video-nystagmography and four laboratory tests (video head impulse test, vestibular-evoked myogenic potentials, caloric testing and audiogram) were used. We applied ten machine learning algorithms to develop classification models. Modelling was performed using three "tiers" of data availability to simulate three clinical settings. "Tier 1" used all available data to simulate the neuro-otology clinic, "Tier 2" used only history, audiogram and caloric test data, representing the general neurology clinic, and "Tier 3" used history alone as occurs in primary care. Model performance was evaluated using tenfold cross-validation.

RESULTS

Data from 160 patients with VM and 114 with MD were used for model development. All models effectively separated the two disorders for all three tiers, with accuracies of 85.77-97.81%. The best performing algorithms (AdaBoost and Random Forest) yielded accuracies of 97.81% (95% CI 95.24-99.60), 94.53% (91.09-99.52%) and 92.34% (92.28-96.76%) for tiers 1, 2 and 3. The best feature subset combination was history, acute video-nystagmography, video head impulse test and caloric testing, and the best single feature subset was history.

CONCLUSIONS

Machine learning models can accurately differentiate between VM and MD and are promising tools to assist diagnosis by medical practitioners with diverse levels of expertise and resources.

Comprehensive Equilibrium Function Tests for an Accurate Diagnosis in Vertigo: A Retrospective Analysis

Background/Objectives: An accurate diagnosis of vertigo is crucial in patient care. Traditional balance function tests often fail to offer independent, conclusive diagnoses. This study aimed to bridge the gap between traditional diagnostic approaches and the evolving landscape of automated diagnostic tools, laying the groundwork for advancements in vertigo care. Methods: A cohort of 1400 individuals with dizziness underwent a battery of equilibrium function tests, and diagnoses were established based on the criteria by the Japanese Society for Vertigo and Equilibrium. A multivariate analysis identified the key diagnostic factors for various vestibudata nlar disorders, including Meniere's disease, vestibular neuritis, and benign paroxysmal positional vertigo. Results: This study underscored the complexity of diagnosing certain disorders such as benign paroxysmal positional vertigo, where clinical symptoms play a crucial role. Additionally, it highlighted the utility of specific physical balance function tests for differentiating central diseases. These findings bolster the reliability of established diagnostic tools, such as audiometry for Meniere's disease and spontaneous nystagmus for vestibular neuritis. Conclusions: This study concluded that a multifaceted approach integrating multiple diagnostic indicators is crucial for accurate clinical decisions in vestibular disorders. Future studies should incorporate novel tests, quantitative assessments, and advanced technologies to enhance the diagnostic capabilities of vestibular medicine.

Neurological update: neuro-otology 2023

Much has changed since our last review of recent advances in neuro-otology 7 years ago. Unfortunately there are still not many practising neuro-otologists, so that most patients with vestibular problems need, in the first instance, to be evaluated and treated by neurologists whose special expertise is not neuro-otology. The areas we consider here are mostly those that almost any neurologist should be able to start managing: acute spontaneous vertigo in the Emergency Room-is it vestibular neuritis or posterior circulation stroke; recurrent spontaneous vertigo in the office-is it vestibular migraine or Meniere's disease and the most common vestibular problem of all-benign positional vertigo. Finally we consider the future: long-term vestibular monitoring and the impact of machine learning on vestibular diagnosis.

Machine learning and bioinformatics analysis to identify autophagy-related biomarkers in peripheral blood for rheumatoid arthritis

Background: Although rheumatoid arthritis (RA) is a common autoimmune disease, the precise pathogenesis of the disease remains unclear. Recent research has unraveled the role of autophagy in the development of RA. This research aims to explore autophagy-related diagnostic biomarkers in the peripheral blood of RA patients. Methods: The gene expression profiles of GSE17755 were retrieved from the gene expression ontology (GEO) database. Differentially expressed autophagy-related genes (DE-ARGs) were identified for the subsequent research by inserting autophagy-related genes and differentially expressed genes (DEGs). Three machine learning algorithms, including random forest, support vector machine recursive feature elimination (SVM-RFE), and least absolute shrinkage and selection operator (LASSO), were employed to identify diagnostic biomarkers. A nomogram model was constructed to assess the diagnostic value of the biomarkers. The CIBERSORT algorithm was performed to investigate the correlation of the diagnostic biomarkers with immune cells and immune factors. Finally, the diagnostic efficacy and differential expression trend of diagnostic biomarkers were validated in multiple cohorts containing different tissues and diseases. Results: In this study, 25 DE-ARGs were identified between RA and healthy individuals. In addition to "macroautophagy" and "autophagy-animal," DE-ARGs were also associated with several types of programmed cell death and immune-related pathways according to GO and KEGG analysis. Three diagnostic biomarkers, EEF2, HSP90AB1 and TNFSF10, were identified by the random forest, SVM-RFE, and LASSO. The nomogram model demonstrated excellent diagnostic value in GSE17755 (AUC = 0.995, 95% CI: 0.988-0.999). Furthermore, immune infiltration analysis showed a remarkable association between EEF2, HSP90AB1, and TNFSF10 expression with various immune cells and immune factors. The three diagnostic biomarkers also exhibited good diagnostic efficacy and demonstrated the same trend of differential expression in multiple validation cohorts. Conclusion: This study identified autophagy-related diagnostic biomarkers based on three machine learning algorithms, providing promising targets for the diagnosis and treatment of RA.

Automated machine learning to predict the difficulty for endoscopic resection of gastric gastrointestinal stromal tumor

Background

Accurate preoperative assessment of surgical difficulty is crucial to the success of the surgery and patient safety. This study aimed to evaluate the difficulty for endoscopic resection (ER) of gastric gastrointestinal stromal tumors (gGISTs) using multiple machine learning (ML) algorithms.

Methods

From December 2010 to December 2022, 555 patients with gGISTs in multi-centers were retrospectively studied and assigned to a training, validation, and test cohort. A difficult case was defined as meeting one of the following criteria: an operative time ≥ 90 min, severe intraoperative bleeding, or conversion to laparoscopic resection. Five types of algorithms were employed in building models, including traditional logistic regression (LR) and automated machine learning (AutoML) analysis (gradient boost machine (GBM), deep neural net (DL), generalized linear model (GLM), and default random forest (DRF)). We assessed the performance of the models using the areas under the receiver operating characteristic curves (AUC), the calibration curve, and the decision curve analysis (DCA) based on LR, as well as feature importance, SHapley Additive exPlanation (SHAP) Plots and Local Interpretable Model Agnostic Explanation (LIME) based on AutoML.

Results

The GBM model outperformed other models with an AUC of 0.894 in the validation and 0.791 in the test cohorts. Furthermore, the GBM model achieved the highest accuracy among these AutoML models, with 0.935 and 0.911 in the validation and test cohorts, respectively. In addition, it was found that tumor size and endoscopists' experience were the most prominent features that significantly impacted the AutoML model's performance in predicting the difficulty for ER of gGISTs.

Conclusion

The AutoML model based on the GBM algorithm can accurately predict the difficulty for ER of gGISTs before surgery.

Artificial intelligence for early stroke diagnosis in acute vestibular syndrome

Objective

Measuring the Vestibular-Ocular-Reflex (VOR) gains with the video head impulse test (vHIT) allows for accurate discrimination between peripheral and central causes of acute vestibular syndrome (AVS). In this study, we sought to investigate whether the accuracy of artificial intelligence (AI) based vestibular stroke classification applied in unprocessed vHIT data is comparable to VOR gain classification.

Methods

We performed a prospective study from July 2015 until April 2020 on all patients presenting at the emergency department (ED) with signs of an AVS. The patients underwent vHIT followed by a delayed MRI, which served as a gold standard for stroke confirmation. The MRI ground truth labels were then applied to train a recurrent neural network (long short-term memory architecture) that used eye- and head velocity time series extracted from the vHIT examinations.

Results

We assessed 57 AVS patients, 39 acute unilateral vestibulopathy patients (AUVP) and 18 stroke patients. The overall sensitivity, specificity and accuracy for detecting stroke with a VOR gain cut-off of 0.57 was 88.8, 92.3, and 91.2%, respectively. The trained neural network was able to classify strokes with a sensitivity of 87.7%, a specificity of 88.4%, and an accuracy of 87.9% based on the unprocessed vHIT data. The accuracy of these two methods was not significantly different (p = 0.09).

Conclusion

AI can accurately diagnose a vestibular stroke by using unprocessed vHIT time series. The quantification of eye- and head movements with the use of machine learning and AI can serve in the future for an automated diagnosis in ED patients with acute dizziness. The application of different neural network architectures can potentially further improve performance and enable direct inference from raw video recordings.

Development and Validation of the Predictive Model for the Differentiation between Vestibular Migraine and Meniere's Disease

(1) Background: Vestibular migraine (VM) and Meniere’s disease (MD) share multiple features in terms of clinical presentations and auditory-vestibular dysfunctions, e.g., vertigo, hearing loss, and headache. Therefore, differentiation between VM and MD is of great significance. (2) Methods: We retrospectively analyzed the medical records of 110 patients with VM and 110 patients with MD. We at first established a regression equation by using logistic regression analysis. Furthermore, sensitivity, specificity, accuracy, positive predicted value (PV), and negative PV of screened parameters were assessed and intuitively displayed by receiver operating characteristic curve (ROC curve). Then, two visualization tools, i.e., nomograph and applet, were established for convenience of clinicians. Furthermore, other patients with VM or MD were recruited to validate the power of the equation by ROC curve and the Gruppo Italiano per la Valutazione degli Interventi in Terapia Intensiva (GiViTI) calibration belt. (3) Results: The clinical manifestations and auditory-vestibular functions could help differentiate VM from MD, including attack frequency (X5), phonophobia (X13), electrocochleogram (ECochG) (X18), head-shaking test (HST) (X23), ocular vestibular evoked myogenic potential (o-VEMP) (X27), and horizontal gain of vestibular autorotation test (VAT) (X30). On the basis of statistically significant parameters screened by Chi-square test and multivariable double logistic regression analysis, we established a regression equation: P = 1/[1 + e^{−(−2.269× X5 − 2.395× X13 + 2.141× X18 + 3.949 × X23 + 2.798× X27 − 4.275× X30(1) − 5.811× X30(2) + 0.873)}] (P, predictive value; e, natural logarithm). Nomographs and applets were used to visualize our result. After validation, the prediction model showed good discriminative power and calibrating power. (4) Conclusions: Our study suggested that a diagnostic algorithm based on available clinical features and an auditory-vestibular function regression equation is clinically effective and feasible as a differentiating tool and could improve the differential diagnosis between VM and MD.

Videooculography “HINTS” in Acute Vestibular Syndrome: A Prospective Study

Objective

A three-step bedside test (“HINTS”: Head Impulse-Nystagmus-Test of Skew), is a well-established way to differentiate peripheral from central causes in patients with acute vestibular syndrome (AVS). Nowadays, the use of videooculography gives physicians the possibility to quantify all eye movements. The goal of this study is to compare the accuracy of VOG “HINTS” (vHINTS) to an expert evaluation.

Methods

We performed a prospective study from July 2015 to April 2020 on all patients presenting at the emergency department with signs of AVS. All the patients underwent clinical HINTS (cHINTS) and vHINTS followed by delayed MRI, which served as a gold standard for stroke confirmation.

Results

We assessed 46 patients with AVS, 35 patients with acute unilateral vestibulopathy, and 11 patients with stroke. The overall accuracy of vHINTS in detecting a central pathology was 94.2% with 100% sensitivity and 88.9% specificity. Experts, however, assessed cHINTS with a lower accuracy of 88.3%, 90.9% sensitivity, and 85.7% specificity. The agreement between clinical and video head impulse tests was good, whereas for nystagmus direction was fair.

Conclusions

vHINTS proved to be very accurate in detecting strokes in patients AVS, with 9% points better sensitivity than the expert. The evaluation of nystagmus direction was the most difficult part of HINTS.