|Abstract:||Cardiac diseases are the leading cause of mortality in the United States, accounting for every one in seven deaths. There are a large proportion of cardiac diseases that need histopathological examination by pathologists for a conclusive diagnosis, but this technique hasn’t been improved upon in the past decade. In this work, we have attempted to advance the current state of histology by developing stainless staining protocols using infrared spectroscopy.
The current gold standard to identify cardiovascular complications such as ischemia, fibrosis, alcoholic cardiomyopathy and transplant rejection is biopsy followed by histology. This approach lacks in many aspects. Major challenges faced by pathologists are: addressing inter-observer variability and experimental variations in stain development, and developing approaches for in-situ histopathology. Specifically, in the case of cardiac transplants, regular monitoring of the transplant is required in order to ensure that the body accepts the transplant. This is done by collecting tissue biopsies at specific time intervals. The presence of lymphocytic infiltration and accompanying fibrosis is indicative of transplant rejection. A prompt clinical action is required if rejection is identified in the biopsy. Cardiac transplant patients can benefit from techniques that can identify lymphocytic infiltration and fibrosis with high accuracy, complementing current pathology practice and giving greater opportunity to pathologists to study complex cases. In the first part of this work, we used infrared spectroscopy coupled with supervised Bayesian classification to identify lymphocytic infiltration and fibrosis in the myocardium in endomyocardial biopsy samples. This classifier was robust and could be easily applied to identify lymphocytes in the tissue and to differentiate between fibrosis in endocardium with fibrosis in myocardium which stains similarly in hematoxylin and eosin stain (H&E).
Repeated biopsy procedures can cause significant trauma to the patient, and often the surgeons require real time histopathology information of the tissue during surgeries. This cannot be accomplished by traditional histology where the tissue sample needs to be excised, sectioned and stained for analysis. Since infrared spectroscopy in stainless, probe-based instruments can be developed to provide detection in-situ but were earlier limited by the speed of imaging using Fourier Transform infrared spectrometers. The problem of speed can be overcome by using quantum cascade laser-based discrete frequency infrared (DFIR) imaging instruments. In the second part of this work, we analyzed data collected on recently developed discrete frequency instruments and compared it to data collected on FT-IR imaging instruments. This was done by unsupervised data clustering to observe histological classes in both types of data.
After establishing that the data collected in DFIR mode retained spectral differences between the histological classes to enable their differentiation, in part three of this work we have done extensive analysis of classification approaches that can be applied to the DFIR data. This study will be relevant to many of the previously built Bayesian classification models that need to be evaluated for their applicability on data collected in discrete frequency mode. In addition, we identified specific spectral features that could be used to differentiate between fibrosis and normal tissue in cardiac biopsy samples computationally at high speed using discrete frequency approach. This can give way to utilization of this model in fiber optic probe-based technology for on-site detection of fibrosis in patients.