Self-calibrating Intelligent OCT-SLO SystemMayank Goswami

Abstract

A unique sample independent 3D self calibration methodology is tested on a unique optical coherence tomography and multi-spectral scanning laser ophthalmoscope (OCT-SLO) hybrid system. Operators visual cognition is replaced by computer vision using the proposed novel fully automatic AI-driven system design. Sample specific automatic contrast adjustment of the beam is achieved on the pre-instructed region of interest. The AI model deduces infrared, fluorescence, and visual spectrum optical alignment by estimating pre-instructed features quantitatively. The tested approach, however, is flexible enough to utilize any apt AI model. Relative comparison with classical signal-to-noise-driven automation is shown to be 200 percent inferior and 130 percent slower than the AI-driven approach. The best spatial resolution of the system is found to be (a) 2.41 microns in glass bead eye phantom, 0.76 with STD 0.46 microns in the mouse retina in the axial direction, and (b) better than 228 line pair per millimeter (lp per mm) or 2 microns for all three spectrums, i.e., 488 nm, 840 nm, and 520 to 550 nm emission in coronal, frontal or x-y plane. Intelligent automation reduces the possibility of developing cold cataracts (especially in mouse imaging) and patient-associated discomfort due to delay during manual alignment by facilitating easy handling for swift ocular imaging and better accuracy. The automatic novel tabletop compact system provides true functional 3D images in three different spectrums for dynamic sample profiles. This is especially useful for photodynamic imaging treatment.