A four class model for digital breast histopathology using high-definition Fourier transform infrared (FT-IR) spectroscopic imaging

Abstract

High-definition (HD) Fourier transform infrared (FT-IR) spectroscopic imaging is an emerging technique that not only enables chemistry-based visualization of tissue constituents, and label free extraction of biochemical information but its higher spatial detail makes it a potentially useful platform to conduct digital pathology. This methodology, along with fast and efficient data analysis, can enable both quantitative and automated pathology. Here we demonstrate a combination of HD FT-IR spectroscopic imaging of breast tissue microarrays (TMAs) with data analysis algorithms to perform histologic analysis. The samples comprise four tissue states, namely hyperplasia, dysplasia, cancerous and normal. We identify various cell types which would act as biomarkers for breast cancer detection and differentiate between them using statistical pattern recognition tools i.e. Random Forest (RF) and Bayesian algorithms. Feature optimization is integrally carried out for the RF algorithm, reducing computation time as well as redundant spectral features. We achieved an order of magnitude reduction in the number of features with comparable prediction accuracy to that of the original feature set. Together, the demonstration of histology and selection of features paves the way for future applications in more complex models and rapid data acquisition.

SIproc: an open-source biomedical data processing platform for large hyperspectral images

“…Our FFT is implemented using the cuFFT GPU accelerated FFT library and performs ≈ 10× faster than those provided with the instrumentation. Other algorithms implemented in SIproc include: spectral baseline correction, normalization, image classification using random forests, 27,30 digital staining, 22 MNF noise reduction, 14 dimension reduction, masking and thresholding, and standard image manipulation tools such as merging and cropping. SIproc also includes a tool for visualization of hyperspectral images using streaming.…”

“…In addition to standard pre-processing and statistical methods, we have also included supervised and unsupervised learning methods such as k-means clustering, random forests, 25–27 and artificial neural networks designed to implement stainless staining. 22 …”

Infrared Spectroscopic Imaging Advances as an Analytical Technology for Biomedical Sciences

“…All this helps in creating robust classification models, but on the down side, the scattering effects are more pronounced in HD and a question arises if those new variability sources will not prevent or diminish the capabilities of machine learning models. Recent studies on prostate 101 and breast 35 cancer tissues have shown that it is not the case and advanced classification tools can handle the increased information content. HD is very likely to become a standard in IR imaging; however, the decreased SNR puts a limit on the speed of acquisition, and new advancements in the technology are required in either broadband sources or new generation of detectors.…”

“…103,252 Breast tissue analysis using HD has also been shown, 253 followed by the creation of a more complex tissue structure classification. 35 HD imaging has also helped in identification of parasites in single erythrocytes 107 by increasing spatial resolution of the parasite signal and in observing tendon damage using linearly polarized light. 106 All these studies point cautiously to the idea that improved computational methods may be able to overcome variability in data that may arise from smaller pixels.…”

Statistical Considerations and Tools to Improve Histopathologic Protocols with Spectroscopic Imaging

“…6 Digital histopathological characterization has also been proposed for a variety of tissue types, including liver, 7,8 lymph node analysis, 9,10 kidney, 11,12 lung adenocarcinoma, 13 brain tumors, [14][15][16] head and neck tumors, 17 oral cavity, 18 prostate, 19,20 bone, [21][22][23] colon, [24][25][26][27][28] and breast. 6,[29][30][31][32][33][34][35] In general, for these varieties of tissue types and diseases, spectroscopic imaging is coupled to a variety of data mining approaches. Recent developments in the capabilities of analytical instruments designed for vibrational IR spectroscopy to resolve microscopic structures have drastically advanced the feasibility of IR spectroscopic imaging for routine histopathology.…”