In clinical practice, sleep stage classification (SSC) is a crucial step for physicians in sleep assessment and sleep disorder diagnosis. However, traditional sleep stage classification relies on manual work by sleep experts, which is time-consuming and labor-intensive. Faced with this obstacle, computer-aided diagnosis (CAD) has the potential to become an intelligent assistant tool for sleep experts, aiding doctors in the assessment and decision-making process. In fact, in recent years, CAD supported by artificial intelligence, especially deep learning (DL) techniques, has been widely applied in SSC. DL offers higher accuracy and lower costs, making a significant impact. In this paper, we will systematically review SSC research based on DL methods (DL-SSC). We explores DL-SSC from several important perspectives, including signal and data representation, data preprocessing, deep learning models, and performance evaluation. Specifically, this paper addresses three main questions: (1) What signals can DL-SSC use? (2) What are the various methods to represent these signals? (3) What are the effective DL models? Through addressing on these questions, this paper will provide a comprehensive overview of DL-SSC.

We report a new fluorescence resonance energy transfer (FRET) sensing platform for the sensitive detection of caspase-3 activity in vitro and in cells using NaGdF₄:Yb³⁺,Er³⁺@NaGdF₄ upconversion nanoparticles (UCNPs) as the energy donor and Rhodamine B (RB) as the energy acceptor. The phosphorylated RB-modified peptide containing a caspase-3 cleavage site and cell-penetrating peptide (CPP) motif (sequence, (RB)-DEVDGGS(p)GCGT(p)GRKKRRQRRRPQ) is immobilized on the UCNP surface via the strong coordination interaction between Gd³⁺ ions with phosphate. After the cleavage of DEVD by caspase-3, the RB is released from the UCNP surface and the reduced upconversion luminescence (UCL) is recovered. Under the optimum conditions, the recovery ratio of the UCL is linearly dependent on the caspase-3 concentration within the range of 0.01 to 1000 pg mL⁻¹ and with a limit of detection (LOD) of 0.01 pg mL⁻¹ (S/N = 3). In particular, the as-proposed UCNP-based FRET sensing platform has reasonable selectivity which is successfully employed to monitor caspase-3 activity in drug-induced apoptosis of HeLa cells.

Accurate and sensitive detection of disease-related microRNAs (miRNAs) is of great significance for early disease diagnosis. In this work, a DNA tetrahedron nanoprobe (DTNP)-based fluorescence resonance energy transfer (FRET) sensing platform (termed DTNP sensor) was constructed for sensitive detection of tumor-related miRNA (e.g., hsa-miR-146b-5p) with DNA assisted cyclic amplification. DTNP was synthesized by DNA self-assembly. In the absence of hsa-miR-146b-5p, the fluorescence DNA (HP) modified with FAM at the 5′ terminal and TAMRA at the 3′ terminal cannot form the hairpin structure because of the hybridization with the extended DNA strand of the DNA tetrahedron, resulting in a low FRET effect. In the presence of hsa-miR-146b-5p, it would complementarily hybridize with the extended DNA strand of the DNA tetrahedron, leading to the release of HP and occurrence of strong FRET. Thus, the concentration of hsa-miR-146b-5p can be revealed by the change in the fluorescence intensity. Moreover, an assistant DNA was employed to replace hsa-miR-146b-5p for cyclic signal amplification, which can further enhance the detection sensitivity. Under the optimal experimental conditions, the limit of detection for hsa-miR-146b-5p was as low as 6 pM (S/N = 3). Furthermore, the DTNP sensor was successfully applied to evaluate the hsa-miR-146b-5p expression levels in different cell lines. The inhibition of hsa-miR-146b-5p expression in different cells was also investigated and a satisfactory result was obtained.