Matlab r2021b

All computations were performed on a consumer-grade laptop (CPU: Intel i7-8750H, RAM: 16 GB DDR4, GPU: 8 GB NVIDIA GeForce RTX 2070 w/ Max-Q Design) running Windows 10 (Microsoft Corporation, Redmond, WA). All scripts, functions, and neural networks were written and trained in MATLAB R2021b (The MathWorks, Inc., Natick, MA). A conglomeration of MATLAB examples, tutorials, and research articles inspired the segmentation, deep learning, and cell tracking architecture of the image analysis pipeline developed in this study^{36 –42} . All code has been made publicly available through GitHub (https://github.com/BloodML)^{28 ,29} and the UH Dataverse Repository (https://dataverse.tdl.org/dataverse/erythrocyte).^{26 ,27} .

For EEG recording, the rats were divided into 3 groups (control, Chry 100 mg/kg, Chry 200 mg/kg, n = 6). To record EEG and EMG, a 7-pin EEG/EMG cable (363–363; P1 Technologies) was inserted into the implanted pedestal, then connected to a commutator. To allow the rats to get accustomed to the tethered environment, each animal was placed in the experimental apparatus for 3 days prior to the recording. EEG/EMG signals were continuously collected and saved at 200 Hz sampling rates through a recording amplifier (AURA24; Grass-Telefactor, West Warwick, RI, USA) and data acquisition software (Twin 4.5.3; Grass Technologies LTM). Sleep state recordings were made twice, one day before saline (control) or Chry ext treatment (100, and 200 mg/kg) and one hour after the last treatment for 7 days, as shown in the timeline in Figure 1b. Considering the circadian pattern of rodents, 6 h (10:00 to 16:00) were exported and used for analysis. The sleep–wake states (wakefulness, REM, and NREM sleep) were scored semi-automatically by 5 s epochs of EEG/EMG signals using the open-source software AccuSleep [29 (link)] written in MATLAB (R2021b, The MathWorks Inc., Natick, MA, USA). All results of the AccuSleep scoring were validated after visual analysis to reduce errors. Representative EEG/EMG waveforms are shown in Figure 1c.

To compare the groups, the ROMs and required time were analyzed using the independent Student t-test, Welch t-test, or Mann–Whitney U test, which were selected based on data normality and variance tests using JMP 16 (SAS Institute Inc., Cary, NC, USA) and calculated with effect size r [26 (link)]. Sequential joint angles were used for statistical parametric mapping (SPM) of two independent Student t-tests [27 (link),28 (link)] to compare joint angle positions with the timing of movements between groups. The advantage of SPM is that entire continuous biomechanical data can be analyzed by calculating test statistics for the null hypothesis. The results can indicate periods of group differences in movement cycles in joint angle data of ADL tasks [9 (link),11 (link),16 (link)]. SPM was conducted using MATLAB R2021b (MathWorks, Inc., Natick, MA, USA) with an open-source code [29 ]. Three constituent null hypotheses were tested in this study; therefore, the statistical significance level was set at 1 - (1–0.05)^1/3 = 0.017. Effect sizes (r) were small at 0.10, medium at 0.30, and large at 0.50 [26 (link)].

MATLAB R2021b (MathWorks) was used for creating images and fitting mathematical models. Box plots were created with the function daboxplot (https://github.com/frank-pk/DataViz).
Comparisons between the negative control and treatment experiment were performed using a linear mixed model in R [version 4.2.2 (https://www.R-project.org/)]. The assumptions of the model were checked by a normal quantile plot of the residual values and a residual dot plot. For time-dependent data (i.e., accumulated metabolite profiles and biofilm features derived from image analysis), differences between the negative control and treatment at each time point are compared to differences at time point 72 (i.e., reference point). For metabolite profiles and qPCR data, P-values were corrected for simultaneous hypotheses according to Sidak (groups of one factor are compared with each group of the other factor). P-values below 0.05 were considered to be statistically significant.