Ahmed S. BaHammam

Summary Accumulating evidence indicates that patients with isolated rapid eye movement sleep behaviour disorder (iRBD), a prodromal stage of synucleinopathies, show abnormal deposition of misfolded alpha‐synuclein (a‐Syn) in peripheral tissues. The clinical utility of testing for a‐Syn in iRBD is unclear. This meta‐analysis focused on the utility of testing for the abnormal a‐Syn phosphorylated at Ser129 (p‐syn) and a‐Syn seeding activity (a‐Syn seed amplification assays [aSyn‐SAA]). Following an electronic database search, 15 studies were included that provided at a minimum data on test positivity in participants with iRBD. Test positivity from cerebrospinal fluid (CSF) was 80% (95% confidence interval [CI] 68–88%, I 2 = 71%) and for skin was 74.8% (95% CI 53.2–88.5%, I 2 = 64%) for aSyn‐SAA and 78.5% (95% CI 70.4–84.9%, I 2 = 14%) for p‐syn. The phenoconversion rate ratio of biopsy‐positive versus biopsy‐negative iRBD was 1.28 (95% CI 0.68–2.41, I 2 = 0%). Skin as a source had a specificity of 99% (95% CI 95–100%, I 2 = 0%; p = 0.01 compared to CSF). As a test, p‐syn, had a specificity of 100% (95% CI 93–100%, I 2 = 0%; p < 0.001) compared to aSyn‐SAA. The odds ratio of a‐Syn test positivity in iRBD versus other RBDs was 112 (95% CI 20–629, I 2 = 0%). These results demonstrate clinically significant test positivity in iRBD and favour skin over CSF as the source of a‐Syn pathological analysis, and p‐syn over aSyn‐SAA as the testing method. Overall, these findings indicate that testing for a‐Syn could help in differentiating iRBD from RBD secondary to other conditions.

Scoring polysomnography for obstructive sleep apnea diagnosis is a laborious, long, and costly process. Machine learning approaches, such as deep neural networks, can reduce scoring time and costs. However, most methods require prior filtering and preprocessing of the raw signal. Our work presents a novel method for diagnosing obstructive sleep apnea using a transformer neural network with learnable positional encoding, which outperforms existing state-of-the-art solutions. This approach has the potential to improve the diagnostic performance of oximetry for obstructive sleep apnea and reduce the time and costs associated with traditional polysomnography. Contrary to existing approaches, our approach performs annotations at one-second granularity. Allowing physicians to interpret the model’s outcome. In addition, we tested different positional encoding designs as the first layer of the model, and the best results were achieved using a learnable positional encoding based on an autoencoder with structural novelty. In addition, we tried different temporal resolutions with various granularity levels from 1 to 360 s. All experiments were carried out on an independent test set from the public OSASUD dataset and showed that our approach outperforms current state-of-the-art solutions with a satisfactory AUC of 0.89, accuracy of 0.80, and F1-score of 0.79.