Andreas Tycho

Machine-learning classification of non-melanoma skin cancers from image features obtained by optical coherence tomography

Background/purpose: A number of publications have suggested that optical coherence tomography (OCT) has the potential for non‐invasive diagnosis of skin cancer. Currently, individual diagnostic features do not appear sufficiently discriminatory. The combined use of several features may however be useful.

Extraction of optical scattering parameters and attenuation compensation in optical coherence tomography images of multilayered tissue structures

A recently developed analytical optical coherence tomography (OCT) model [Thrane et al., J. Opt. Soc. Am. A 17, 484 (2000)] allows the extraction of optical scattering parameters from OCT images, thereby permitting attenuation compensation in those images. By expanding this theoretical model, we have developed a new method for extracting optical scattering parameters from multilayered tissue structures in vivo. To verify this, we used a Monte Carlo (MC) OCT model as a numerical phantom to simulate the OCT signal for heterogeneous multilayered tissue. Excellent agreement between the extracted values of the optical scattering properties of the different layers and the corresponding input reference values of the MC simulation was obtained, which demonstrates the feasibility of the method for in vivo applications. This is to our knowledge the first time such verification has been obtained, and the results hold promise for expanding the functional imaging capabilities of OCT.

Advanced modelling of optical coherence tomography systems

Analytical and numerical models for describing and understanding the light propagation in samples imaged by optical coherence tomography (OCT) systems are presented. An analytical model for calculating the OCT signal based on the extended Huygens-Fresnel principle valid both for the single and multiple scattering regimes is reviewed. An advanced Monte Carlo model for calculating the OCT signal is also reviewed, and the validity of this model is shown through a mathematical proof based on the extended Huygens-Fresnel principle. Moreover, for the first time the model is verified experimentally. From the analytical model, an algorithm for enhancing OCT images is developed: the so-called true-reflection algorithm in which the OCT signal may be corrected for the attenuation caused by scattering. For the first time, the algorithm is demonstrated by using the Monte Carlo model as a numerical tissue phantom. Such algorithm holds promise for improving OCT imagery and to extend the possibility for functional imaging.

Comment on "Excitation with a focused, pulsed optical beam in scattering media: diffraction effects".

To incorporate the wave properties of light, it was recently proposed to modify the Monte Carlo-based photon transport model to a semi-quantum-mechanical representation by considering each ray as a photon wave packet [Appl. Opt. 39, 5244 (2000)]. It is not clear from the paper whether each photon wave packet is considered representative of an entire plane wave or if each is spatially localized. However, for each interpretation we identify problems with the approach suggested for combining the wave-packet contributions. These include violation of the principle of conservation of energy and the use of a scattering phase function that is incompatible with the suggested way of calculating the intensity values. These issues render the approach impractical.

Modeling the optical coherence tomography geometry using the extended Huygens-Fresnel principle and Monte Carlo simulations

We review a new theoretical description of the optical coherence tomography (OCT) geometry for imaging in highly scattering tissue. The new model is based on the extended Huygens-Fresnel principle, and it is valid in the single and multiple scattering regimes. Furthermore, we simulate the operation of the OCT system using a specially adapted Monte Carlo simulation code. To enable Monte Carlo simulation of the coherent mixing of the sample and reference beams the code uses a method of calculating the OCT signal derived using the extended Huygens-Fresnel principle. Results obtained with the Monte Carlo simulation and the new theoretical description compare favorably. Finally, the application of the extended Huygens-Fresnel principle for extracting optical scattering properties is used to obtain a so-called true reflection algorithm.

Optical coherence tomography in clinical examination of non-pigmented skin malignancies

Optical coherence tomography (OCT) images of basal cell carcinomas (BCCs) have been acquired using a compact handheld probe with an integrated video camera allowing the OCT images to be correlated to a skin surface image. In general the healthy tissue of the skin has an obvious stratified structure, whereas the cancerous tissue shows a more homogenous structure. Thus it was demonstrated that it is possible to distinguish BCCs from healthy tissue by means of OCT. Furthermore different histological types of BCC were identified. Comparison of OCT images taken prior to and immediately after photodynamic therapy clearly shows the tissue response to the treatment, and indicates local oedema in the treated area.

Focusing problem in OCT: comparison of Monte-Carlo simulations, the extended Huygens-Fresnel principle, and experiments

In the later years, a great effort has ben put into simulation of the geometry of an optical coherence tomography (OCT) system. Recently, a new analytical model of the OCT geometry has been developed based on the extended Huygens-Fresnel (EHF) principle. Although advanced, the result of the model are surprisingly simple and easy to handle for e.g. system optimization. To validate this model, new features have been added to the Monte Carlo (MC) simulation program MCML, which is widely used and recognized for its credibility. We have incorporated the true shape of a focused Gaussian beam including the finite size of the beam waist, which previously has been approximated by a point. This enables us to do high-resolution comparison of the intensity distribution in the focus plane and excellent agreement is found between the EHF model and the MC simulations. Results are also compared with previously published modeling result and it is shown that there are substantial differences. We emphasize the importance of the so-called shower curtain effect (SCE), which is an inherent - but often overlooked - effect in light propagation through random media. Finally, we calculate the OCT signal using MC simulation. This is done by keeping track of the path length traveled by each photon packet and restricting its access back into the OCT system from the sample using the antenna theorem. The degradation of the detected signal due to scattering is determined, and compared with the EHF model and experiments. The comparison of MC simulations with EHF allows us to show that the SCE is an inherent effect in MC simulation, and that for common tissue parameters, the EHF model yields the same results as the MC simulation but with faster computation time and with field and phase information available.

Optical coherence tomography in clinical examinations of nonpigmented skin malignancies

Optical coherence tomography (OCT) images of basal cell carcinomas (BCCs) have been acquired using a compact handheld proble with an integrated video camera allowing the OCT images to be correlated to a skin surface image. In general the healthy tissue of the skin has an obvious stratified structure, whereas the cancerous tissue shows a more homogeneous structure. Thus it was demonstrated that it is possible to distinguish BCCs from healthy tissue by means of OCT. Furthermore different histological types of BCC were identified. Comparison of OCT images taken prior to and immediately after photodynamic theory clearly shows the tissue response to the treatment, and indicates local oedema in the treated area.

Extraction of tissue optical properties from optical coherence tomography images for diagnostic purposes (Invited Paper)

The concept of optical coherence tomography (OCT) for high-resolution imaging of tissues in vivo is introduced. Analytical and numerical models for describing and understanding the light propagation in samples imaged by OCT systems are presented. An analytical model for calculating the OCT signal based on the extended Huygens-Fresnel principle and valid both for single and multiple scattering regimes is outlined. From this model, an algorithm for extracting tissue optical properties for multi-layered tissues is derived. The algorithm is first verified for various optical properties and geometries using solid phantoms and numerical simulations. The applicability of the algorithm for extraction of tissue optical properties is then demonstrated for vascular tissue samples ex vivo. With the use of data from numerical phantoms, the validity of the OCT extraction algorithm for a two-layer geometry is further supported. It is concluded that by using optical properties extracted from OCT images of human tissues, the clinical utility of OCT imaging can be substantially increased.

Monte Carlo modeling of optical coherence tomography systems

A Monte Carlo model of light scattering in tissue is described and used to estimate the signal-to-noise-ratio that can be obtained with an optical coherence tomography (OCT) system. By utilizing the correspondence between the Wigner phase-space distribution and the specific intensity in the small-angle approximation, the novel Monte Carlo model is valid for light reflected both from and outside the focal plane of the system. The model is compared with experiments and an analytical model and good agreement is found. It is expected that the model can be used to examine how multiple scattering affects the axial resolution for a given source bandwidth.