Person:

Bouma, Brett

Spectrally encoded confocal microscopy and optical frequency domain imaging are two non-contact optical imaging technologies that provide images of tissue cellular and architectural morphology, which are both used for histopathological diagnosis. Although spectrally encoded confocal microscopy has better transverse resolution than optical frequency domain imaging, optical frequency domain imaging can penetrate deeper into tissues, which potentially enables the visualization of different morphologic features. We have developed a co-registered spectrally encoded confocal microscopy and optical frequency domain imaging system and have obtained preliminary images from human oesophageal biopsy samples to compare the capabilities of these imaging techniques for diagnosing oesophageal pathology.

Hypertrophic scars remain a major clinical problem in the rehabilitation of burn survivors and lead to physical, aesthetic, functional, psychological, and social stresses. Prediction of healing outcome and scar formation is critical for deciding on the best treatment plan. Both subjective and objective scales have been devised to assess scar severity. Whereas scales of the first type preclude cross-comparison between observers, those of the second type are based on imaging modalities that either lack the ability to image individual layers of the scar or only provide very limited fields of view. To overcome these deficiencies, this work aimed at developing a predictive model of scar formation based on polarization sensitive optical frequency domain imaging (PS-OFDI), which offers comprehensive subsurface imaging. We report on a linear regression model that predicts the size of a scar 6 months after third-degree burn injuries in rats based on early post-injury PS-OFDI and measurements of scar area. When predicting the scar area at month 6 based on the homogeneity and the degree of polarization (DOP), which are signatures derived from the PS-OFDI signal, together with the scar area measured at months 2 and 3, we achieved predictions with a Pearson coefficient of 0.57 (p < 10−4) and a Spearman coefficient of 0.66 (p < 10−5), which were significant in comparison to prediction models trained on randomly shuffled data. As the model in this study was developed on the rat burn model, the methodology can be used in larger studies that are more relevant to humans; however, the actual model inferred herein is not translatable. Nevertheless, our analysis and modeling methodology can be extended to perform larger wound healing studies in different contexts. This study opens new possibilities for quantitative and objective assessment of scar severity that could help to determine the optimal course of therapy.