Abstract
Purpose :
To develop a fully automated OCT system to advance research using ex vivo tissue culture, especially for applications requiring high-throughput screening.
Methods :
Both hardware integration (3-axis motorized platform with webcam, Fig. 1) and algorithm innovations (deep learning object detection and tissue segmentation) were made to achieve automated and successive OCT scan acquisition, as well as automated and reliable parameter readout. The pipeline of the OCT-based tissue screening system is summarized in Fig. 2. Briefly, after placing the tissue culture plate in the tray, a photo of the well will be taken by the webcam and examined by an object detection algorithm to identify the existence of tissue sample and determine the precise locations of the sample if any. The coordinates are used to align the OCT light beam automatically in lateral direction. The axial alignment is determined by move the Z- motor and examining the average intensity of the entire B-scan. The position of Z- motor with maximal brightness will be noted and referred later to trigger the saving module for recording. After scan acquisition, the system is allowed a dedicated timeslot for reliable segmentation and readout by a pre-trained deep learning network with a hybrid architecture of Resnet and Multi-scale hierarchical transformer.
Results :
The system demonstrated high repeatability (volume: 0.005 mm3, area: 0.039 mm2, thickness: 0.4 µm) and reproducibility (volume: 0.107 mm3, area: 0.755 mm2 for area, thickness: 12.1 µm) for reliable measurements, as validated using retinal explant from RhoP23H/+ mice. Two metrics, the Z prime factor (Z’) and strictly standardized mean difference (SSMD, denoted as β) that are commonly used to assess assay quality for drug screening, were determined by quantifying the negative and positive treatment groups. The Z’-factors were calculated as 0.70 for volume and 0.55 for thickness measurements. The β value were calculated as 8.8 and 9.3 for volume and thickness respectively. The values indicate en excellent performance of the system for drug screening.
Conclusions :
A fully automated OCT system for ex vivo tissue imaging was successfully developed and validated. We hope our system can fill the gap of OCT for high-throughput tissue screening.
This abstract was presented at the 2024 ARVO Imaging in the Eye Conference, held in Seattle, WA, May 4, 2024.