Oral Cancer Discrimination and Novel Oral Epithelial Dysplasia Stratification Using FTIR Imaging and Machine Learning

Abstract

The Fourier transform infrared (FTIR) imaging technique was used in a transmission model for the evaluation of twelve oral hyperkeratosis (HK), eleven oral epithelial dysplasia (OED), and eleven oral squamous cell carcinoma (OSCC) biopsy samples in the fingerprint region of 1800–950 cm−1. A series of 100 µm × 100 µm FTIR imaging areas were defined in each sample section in reference to the hematoxylin and eosin staining image of an adjacent section of the same sample. After outlier removal, signal preprocessing, and cluster analysis, a representative spectrum was generated for only the epithelial tissue in each area. Two representative spectra were selected from each sample to reflect intra-sample heterogeneity, which resulted in a total of 68 representative spectra from 34 samples for further analysis. Exploratory analyses using Principal component analysis and hierarchical cluster analysis showed good separation between the HK and OSCC spectra and overlaps of OED spectra with either HK or OSCC spectra. Three machine learning discriminant models based on partial least squares discriminant analysis (PLSDA), support vector machines discriminant analysis (SVMDA), and extreme gradient boosting discriminant analysis (XGBDA) were trained using 46 representative spectra from 12 HK and 11 OSCC samples. The PLSDA model achieved 100% sensitivity and 100% specificity, while both SVM and XGBDA models generated 95% sensitivity and 96% specificity, respectively. The PLSDA discriminant model was further used to classify the 11 OED samples into HK-grade (6), OSCC-grade (4), or borderline case (1) based on their FTIR spectral similarity to either HK or OSCC cases, providing a potential risk stratification strategy for the precancerous OED samples. The results of the current study support the application of the FTIR-machine learning technique in early oral cancer detection.

Rapid and Label-Free Histopathology of Oral Lesions Using Deep Learning Applied to Optical and Infrared Spectroscopic Imaging Data

“…Although the DFIR spectral images used in this manuscript do not spatially resolve the nuclei, careful inspection of the darkfield visible images suggests nuclear features. Spectral differences at 1238 cm − 1 were used in the literature to discriminate between hyperplasia, epithelia dysplasia, and oral squamous cell carcinoma [ 18 ]. The prior, FTIR microscopy, study found that the average spectral intensity at 1240 cm − 1 increased as diagnosis became more severe (hyperplasia < dysplasia < squamous cell carcinoma) [ 18 ].…”

“…IR imaging is a powerful tool for studying the spatial variation of the biochemical and molecular structure of tissues without the need for external dyes or reagents that detect molecular patterns [ 15 , 16 ]. The bulk of prior IR spectroscopic imaging data, including for oral cancer [ 17 , 18 , 19 ], has previously been acquired using Fourier transform infrared (FT-IR) microscopy that provides full IR spectral data for all pixels imaged [ 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 ]. Although FT-IR microscopy provides high-dimensional data due to the large spectral bandwidth, a large portion of this bandwidth does not contain biologically relevant vibrations; for example, the “cell-silent” region from ~1900 to 2700 cm − 1 is devoid of biochemical features and increasing the number of spectral features for histopathologic classification is well-known to provide diminishing returns [ 29 ].…”

Leveraging mid-infrared spectroscopic imaging and deep learning for tissue subtype classification in ovarian cancer

“…Prior work often utilizes pixels within the same set of tissue cores during classification. 67,68 This can lead to misleading results since the machine learning algorithm may learn features that correspond to specific patient traits that are challenging to generalize beyond the current dataset. We perform training and validation on mutually exclusive patient cores, achieving robust, generalizable results that enhance scientific rigor and reproducibility.…”

Label-Free Proteomics of Oral Mucosa Tissue to Identify Potential Biomarkers That Can Flag Predilection of Precancerous Lesions to Oral Cell Carcinoma: A Preliminary Study

“…The current gold standard for diagnosis of premalignant or malignant lesions is histopathological evaluation to determine the degree of severity of suspicious lesions to assess the transformation risk of oral premalignant lesions [ 13 ]. Recently, Fourier transform infrared (FTIR) imaging integrated with machine learning found biochemical differences in addition to existing histopathological diagnostic processes and provide oral cancer discrimination and a novel oral epithelial dysplasia stratification strategy identifying heterogeneity between intra-samples and inter-samples of the oral epithelium [ 14 ].…”