Options
Real-time photon-noise limited optical coherence tomography based on pixel-level analog signal processing
Auteur(s)
Beer, Stephan
Editeur(s)
Seitz, Peter
Date de parution
2006
Résumé
This thesis presents the development of a CMOS smart imager for real-time optical coherency tomography (OCT). OCT, a measuring technique based on low-coherence interferometry, allows the acquisition of three-dimensional pictures of discontinuities in the refractive index of the sample. Due to the achieved depth resolution in the micrometer range, it has gained a lot of impact not only for many biomedical but also for industrial applications over the last few years. OCT allows the determination of surfaces as well as the acquisition of tomographic images of transparent and turbid objects with an outstanding depth resolution in the sub-micrometer scale over a depth range of up to several millimeters. The goal of this thesis was the realization of a very fast, robust 3D OCT system performing at close to the physical limit. A multi-channel approach has been chosen and a CMOS smart imager has been developed as they key component of the total system that includes an appropriate electrical, optical, and mechanical system. This system proves that by the use of parallel optical coherence tomography (pOCT) in combination with custom designed CMOS image sensors, real-time imaging with compact devices is realizable. For time domain (TD-) OCT, the reference path length of the interferometer is varied and a simultaneous amplitude demodulation of the optical signal generates the depth image. The cross-sectional or volumetric image of a conventional TD-OCT system is generated by combining many laterally shifted one-dimensional depth scans throughout the object. Therefore, very fast reference path scanners and one or two translation stages are required. This results in a non-negligible mechanical system complexity. In pOCT, the translation scanners are replaced by image-forming optics and a two-dimensional image sensor. This step reduces the depth scan speed greatly and renders 3D image acquisition in some tens of milliseconds possible. On the other hand, the image sensor requirements are extremely challenging: In real-time pOCT, Doppler frequencies of typically 50 kHz and more occur. Image sensors with a frame-rate in the order of 200’000 images per second are required, which are not available today. Therefore, and for other reasons, a custom-designed image sensor is required for pOCT in real-time. CMOS technology allows the production of image sensors with extended functionality that integrates signal processing even at the pixel-level of so-called smart imagers. The pOCT sensor presented in this thesis exploits this CMOS property: Each pixel contains a signal demodulation unit, thereby reducing the necessary external frame-rate by almost two orders of magnitude. The electronic circuits in the pixel are based on the switched current (SI) and switched capacitor (SC) techniques. They fulfill the requirements of the massive parallelism, particularly low power dissipation and small area consumption. By reading out two signals per pixel, not only the modulation amplitude but also its phase is detected. Therefore high-resolution phase topography measurements and polarization-sensitive OCT is also possible. The circuit versatility allows for additional working modes such as e.g. an intensity mode for conventional grey-level imaging for sample placement. The presented pOCT system, which has a measured sensitivity of more than 90 db, works very close to the physical limit imposed by the illumination and the depth scan speed of 20 mm/s. A measured phase accuracy of 2° was achieved in the laboratory. The developed smart imager is the world’s first pOCT image sensor that acquires the demodulation amplitude and phase close to the physical limits in real time. To our knowledge, the world’s fastest TD-OCT system was realized based on this sensor.
Notes
Thèse de doctorat : Université de Neuchâtel, 2006 ; 1871
Identifiants
Type de publication
doctoral thesis
Dossier(s) à télécharger
