Label-free classification of colon cancer grading using infrared spectral histopathology

Abstract

In recent years spectral histopathology (SHP) has been established as a label-free method to identify cancer within tissue. Herein, this approach is extended. It is not only used to identify tumour tissue with a sensitivity of 94% and a specificity of 100%, but in addition the tumour grading is determined. Grading is a measure of how much the tumour cells differ from the healthy cells. The grading ranges from G1 (well-differentiated), to G2 (moderately differentiated), G3 (poorly differentiated) and in rare cases to G4 (anaplastic). The grading is prognostic and is needed for the therapeutic decision of the clinician. The presented results show good agreement between the annotation by SHP and by pathologists. A correlation matrix is presented, and the results show that SHP provides prognostic values in colon cancer, which are obtained in a label-free and automated manner. It might become an important automated diagnostic tool at the bedside in precision medicine.

Spatial and molecular resolution of diffuse malignant mesothelioma heterogeneity by integrating label-free FTIR imaging, laser capture microdissection and proteomics

“…The patients are all male between 36 and 83 years old with a DMM in the pleura parietalis or visceralis. Here, the sample set is actually small because mesothelioma is a rare and heterogeneous disease13, but our previous work has shown that the method is very robust, even with such small sample number, due to the high number of spectral markers per sample (~10 Mio) high specificity and sensitivity is obtained343536.…”

“…We have presented success with FTIR imaging tissue annotation in previous studies on the automated classification of colorectal cancer (accuracy of 96%, sensitivity of 94%, and specificity of 100%)34 and the differential diagnosis of the cancer, as well as its grading resulting in high accuracy, sensitivity, and specificity (G1: 99%, 94%, 100%; G2: 86%, 83%, 94%; G3: 87%, 89%, 87%)35. Furthermore, we analysed lung tumours and thoracal tumours, showing that differential diagnosis down to the subtype of adenocarcinomas of the lung is possible using FTIR imaging (accuracy of 96%; no binary classification so sensitivity and specificity cannot be calculated)36.…”

Automated classification of hepatocellular carcinoma differentiation using multiphoton microscopy and deep learning

“…In addition, although the collaborating pathologist reviewed nearly 5 years of archives, with a total of approximately 600 samples, no cases were found diagnosed as undifferentiated carcinoma; that is to say, in this work we have no way to pre‐screen G4. For the appearance of rare G4 , we propose the following strategies: According to the grading systems [https://www.cancer.gov/about-cancer/diagnosis-staging/prognosis/tumor-grade-fact-sheet], G3 and G4 belong to high grade with similar image features. Ideally, G4 is fed into the existing CNN, and the feedback should be G3.…”

Mid‐infrared spectral microimaging of inflammatory skin lesions

“…In this study, acne, rosacea and psoriasis samples, representing 3 key inflammatory skin diseases, were investigated. By combining IR spectroscopy with multivariate statistical techniques of multidimensional data processing, it is possible to develop a biophotonic approach for the identification and characterization of different tissue structures . This approach exploits, at micron level resolution, the non‐destructive feature and the access to molecular information provided by spectral imaging.…”