Labview 2015

Large FoV images were acquired with the JM-OCT prototype in combination with a motorized translation stage (MLS203-1, Thorlabs) moving in $x$ and $y$ direction. The stage was integrated into the sample arm and was able to scan large field-of-views up to several square centimeter. The maximal range of the stage was 11.0 cm $\times <!--\; \times \; -->$ 7.5 cm. A custom-made LabView (LabView 2015, Version 15.0, 64-bit, National Instruments) program controlled the movement of the translation stage [59 (link)]. The OCT data were acquired with an overlap, in the lateral direction, of 5% and after post-processing the volumes were stitched together using the Pairwise stitching plugin of Fiji [60 (link)].

All stored data were exported from ICM+ software to plain text files and imported in MATLAB (Release 2019b, The MathWorks, Inc., Natick, MA, USA). NIRS data was first visually inspected, and artifacts were removed (details on the applied methodology can be found in Supplementary Figure S1). Then, the data was stored in ten-minute data segments per patient and imported into a custom made LabVIEW program (LabVIEW 2015, National Instruments, Austin, TX, USA) calculating the CA-estimates (Section 2.4.1) as used in Elting et al. [9 (link)]. Sufficiently slow ABP oscillations are required for a reliable CA assessment [10 (link)]. The amount of slow ABP and resulting slow oxyHb and deoxyHb oscillations were quantified by the Power Spectral Density (PSD).

The robotic exoskeleton used in this study (Figure 1) consisted of active hip and knee joints and passive ankle joints. Active joints were powered by electrical motors with a gear ratio of 76 : 1, with assistive torque delivered to the wearer's joints in the sagittal plane. Active joints adopted a series elastic actuation mechanism [21 (link), 22 (link)], enabling accurate control of the interaction torque between robot and human joints. Dynamic ankle-foot orthosis was used as the passive ankle joint to support the weight of the exoskeleton and to provide a sufficient degree of freedom in the wearer's ankle joint. Foot pressure under the metatarsal joint and the heel was measured using silicon tubes and air pressure sensors to estimate the ground reaction force [23 (link)]. All algorithms required to operate the exoskeleton, such as sensor signal processing and the control algorithm, were implemented using the embedded control board (sbrio 9651, National Instruments, TX, USA) and software (LabVIEW 2015, National Instruments, TX, USA). The embedded control board and battery, enclosed in a backpack, enabled the exoskeleton to be fully mobile. The total weight of the robotic exoskeleton was 13 kg.

Fluorescence images consisted of 500 (x) × 500 (y) pixels and the processed OCM data sets comprised 500 (x) × 500 (y) × 4096 (z) pixels, each covering the same 200 × 200 µm2 lateral FOV. OCM volumes and FL intensities were acquired in 8.3 seconds using custom software (LabVIEW 2015, Version 15.0, 64-bit, National Instruments, Austin, TX, USA). Data were processed as previously described by Lichtenegger et al. [30 (link)]. After image acquisition, surface flattening was performed and attenuation maps were computed from OCM reflectivity data [65 (link),66 (link)]. Sub-volumes consisting of 100 B-scans were chosen in a manually defined region of interest and average attenuation coefficients were calculated. Respective fluorescence intensities on the tissue surface were quantified and normalized linearly between zero and one, with the upper and lower values defined as the minimum and maximum, respectively, of all data sets. For visualization purposes, Fiji [67 (link)] was used to generate averaged en-face projections and composition images.