Fig 1.
The basic Scanning Laser Optical Tomography (SLOT) setup [
12]. A laser beam is adjusted in size, according to the imaging parameters. The beam is scanned and focused across a sample. The transmitted intensity of the laser is measured for distinct points across the sample, resulting in a transmission image of the sample. Fluorescence can be measured simultaneously. The sample is rotated in discrete angles around 360 degrees to acquire a dataset which can be reconstructed to a tomographic dataset.
Fig 2.
The dependency on the beam parameter and the sample size of a SLOT acquisition.
The beam focus length is determined by the diameter of the sample, such that the focus length (represented by twice the Rayleigh length zR) must cover the sample size d.
Fig 3.
The manual sample alignment process.
(a): The embedded sample is mounted on a kinematic mount. Due to a possible misalignment of the sample within the embedding material, the sample would wobble around the rotation axis. To gain a sharp image, the focus length must cover the entire wobble movement range of the sample, and meaning the focus length must be larger than the sample diameter. (b): After the alignment of the sample to the rotation axis of the mount, the focus length is shortened, compared to (a), to cover the sample’s diameter. This leads to a higher optical resolution.
Fig 4.
The principle of the new virtual alignment method.
(a): The path of the non-aligned, wobbling sample is detected in a fast-pass SLOT acquisition with fewer angle projections. The images are processed with computer vision to calculate a movement prediction. (b): With the calculated movement of the sample, the focus can be shifted, so that it always covers the sample diameter, following the sample position.
Fig 5.
The process chain of the sample movement calculation.
First, the images are processed with a set of computer vision algorithms to detect the sample. Next, the sample positions are calculated and fitted into a sinusoidal movement equation. This equation finally represents the movement of the sample during its non-aligned rotation.
Fig 6.
The process chain of the de-jitter algorithm.
First the sample is detected and segmented. Next the object properties are extracted, whereas the geometric centroid is stored as the object’s movement coordinate. The tracked coordinates are then fitted to a sinusoidal curvature. The images are finally shifted in order to fit to the sinusoidal curve.
Fig 7.
The embedded cell spheroid, used as a test target.
(a): Image of the embedded sample in the SLOT during acquisition. The sample is embedded in a polymer cylinder and placed in a cuvette filled with refractive index matched silicone oil. The green laser scans across the sample. (b): The full field of view image acquired with the transmission channel of SLOT. (c): The used cell spheroid, consisting of 100.000 human osteoblast cells. The sample is located in the red square of (b).
Fig 8.
The setup for the sample tracking algorithm.
(a): An overview of the field of view used in the tracking algorithm. The red rectangle shows the zoomed field of view that was necessary due to the small sample size. (b): The autofluorescence signal in the photo multiplier channel of the zoomed field of view. The cell spheroid emits a detectable signal between 570 and 586 nm. (c): The transmission signal from the photodiode of the zoomed field of view. The spheroid traveled 1920 µm in diameter around the rotation axis.
Fig 9.
Movement of the sample during the rotation with a fixed focus position.
(a): The sample is in focus. The focus length is set to the sample diameter. The image shows the sample on the farthest left position. (b): After a 90-degree rotation, the sample has moved the furthest from the initial focus point and is in the center of the image. The sample appears blurry because it is no longer in focus. (c): After a total of 180-degree rotation, the sample is in focus again and is now on the opposite side of the rotation axis from the initial position in (a). (d): The principle of the movement of the sample and the focus position of the laser is shown in the schematic drawing.
Fig 10.
Excerpts from the raw acquired data set with the tracking algorithm.
(a)-(c): The spheroid shown in different angles. The image retains its sharpness in all angles, which means that the focal length of the laser was successfully shifted with the movement of the sample. The sample position has an offset to the left side of the image in all positions. The background of the data differs from angle to angle, caused by the path of the light through the sample embedding.
Fig 11.
A sinogram of the test sample before and after the dejitter algorithm.
The black sinusoidal line represents the dark spot near the z-center of the sample seen in the transmission projections (Figs 7-10). (a): The uncorrected raw data, acquired with the tracking algorithm. The sinusoidal lines appear jittery due to the irregular wobble movements of the rotation stage. (b): The dejittered dataset. The sinusoidal lines appear smoother and have a constant intensity. The size of the x-shift per line can be seen at the edge of the image as dark pixels.
Fig 12.
The reconstruction of the sinogram of the same image slice as shown in
Fig 11. The slice was reconstructed with the filtered back-projection algorithm. (a): The tomogram of the uncorrected dataset. The structures are blurred and show unsharp edges and doubling artifacts. (b) The tomogram of the jitter corrected dataset. Here, the edges of the sample are sharper and the reconstruction artifacts are reduced. The quality of the image has improved.