Development of a hierarchical double application of crisp cluster validity indices: a proof-of-concept study for automated FTIR spectral histology

Deep learning for the prediction of the chemotherapy response of metastatic colorectal cancer: comparing and combining H&E staining histopathology and infrared spectral histopathology

Colorectal cancer is a global public health problem with one of the highest death rates. It is the second most deadly type of cancer and the third most frequently diagnosed in the world. The present study focused on metastatic colorectal cancer (mCRC) patients who had been treated with chemotherapy-based regimen for which it remains uncertainty about the efficacy for all eligible patients. This is a major problem, as it is not yet possible to test different therapies in view of the consequences on the health of the patients and the risk of progression. Here, we propose a method to predict the efficacy of an anticancer treatment in an individualized way, using a deep learning model constructed on the retrospective analysis of the primary tumor of several patients. Histological sections from tumors were imaged by standard hematoxylin and eosin (HE) staining and infrared spectroscopy (IR). Images obtained were then processed by a convolutional neural network (CNN) to extract features and correlate them with the subsequent progression-free survival (PFS) of each patient. Separately, HE and IR imaging resulted in a PFS prediction with an error of 6.6 and 6.3 months respectively (28% and 26% of the average PFS). Combining both modalities allowed to decrease the error to 5.0 months (21%). The inflammatory state of the stroma seemed to be one of the main features detected by the CNN. Our pilot study suggests that multimodal imaging analyzed with deep learning methods allow to give an indication of the effectiveness of a treatment when choosing.

Live-cell imaging in human colonic monolayers reveals ERK waves limit the stem cell compartment to maintain epithelial homeostasis

The establishment and maintenance of different cellular compartments in tissues is a universal requirement across all metazoans. Maintaining the correct ratio of cell types in time and space allows tissues to form patterned compartments and perform complex functions. Patterning is especially evident in the human colon, where tissue homeostasis is maintained by stem cells in crypt structures that balance proliferation and differentiation. Here, we developed a human 2D patient derived organoid screening platform to study tissue patterning and kinase pathway dynamics in single cells. Using this system, we discovered that waves of ERK signaling induced by apoptotic cells play a critical role in maintaining tissue patterning and homeostasis. If ERK is activated acutely across all cells instead of in wave-like patterns, then tissue patterning and stem cells are lost. Conversely, if ERK activity is inhibited, then stem cells become unrestricted and expand dramatically. This work demonstrates that the colonic epithelium requires coordinated ERK signaling dynamics to maintain patterning and tissue homeostasis. Our work reveals how ERK can antagonize stem cells while supporting cell replacement and the function of the gut.

Automatic Identification of Paraffin Pixels on FTIR Images Acquired on FFPE Human Samples

The transfer of mid-infrared spectral histopathology to the clinic will be possible provided that its application in clinical practice is simple. Rapid analysis of formalin-fixed paraffin-embedded (FFPE) tissue section is thus a prerequisite. The chemical dewaxing of these samples before image acquisition used by the majority of studies is in contradiction with this principle. Fortunately, the in silico analysis of the images acquired on FFPE samples is possible using extended multiplicative signal correction (EMSC). However, the removal of pure paraffin pixels is essential to perform a relevant classification of tissue spectra. So far, this task was possible only if using manual and subjective histogram analysis. In this article, we thus propose a new automatic and multivariate methodology based on the analysis of optimized combinations of EMSC regression coefficients by validity indices and KMeans clustering to separate paraffin and tissue pixels. The validation of our method is performed using simulated infrared spectral images by measuring the Jaccard index between our partitions and the image model, with values always over 0.90 for diverse baseline complexity and signal-to-noise ratio. These encouraging results were also validated on real images by comparing our method with classical ones and by computing the Jaccard index between our partitions and the KMeans partitions obtained on the infrared image acquired on the same samples but after chemical dewaxing, with values always over 0.84.

Towards normalization selection of Raman data in the context of protein glycation: application of validity indices to PCA processed spectra

Vibrational data of biological samples require appropriate pre-processing for ensuring relevant interpretation. Here, mathematical criteria (validity indices) are used to select how to normalize Raman data collected in the protein glycation context.

New insights into spectral histopathology: infrared-based scoring of tumour aggressiveness of squamous cell lung carcinomas

Spectral histopathology, based on infrared interrogation of tissue sections, proved a promising tool for helping pathologists in characterizing histological structures in a quantitative and automatic manner.

Spectral histopathology, based on infrared interrogation of tissue sections, proved a promising tool for helping pathologists in characterizing histological structures in a quantitative and automatic manner. In cancer diagnosis, the use of chemometric methods permits establishing numerical models able to detect cancer cells and to characterize their tissular environment. In this study, we focused on exploiting multivariate infrared data to score the tumour aggressiveness in preneoplastic lesions and squamous cell lung carcinomas. These lesions present a wide range of aggressive phenotypes; it is also possible to encounter cases with various degrees of aggressiveness within the same lesion. Implementing an infrared-based approach for a more precise histological determination of the tumour aggressiveness should arouse interest among pathologists with direct benefits for patient care. In this study, the methodology was developed from a set of samples including all degrees of tumour aggressiveness and by constructing a chain of data processing steps for an automated analysis of tissues currently manipulated in routine histopathology.

Digital de-waxing on FTIR images

This paper presents a procedure that digitally neutralizes the contribution of paraffin to FTIR hyperspectral images. A brief mathematical derivation of the procedure is demonstrated and applied on one normal human colon sample to exemplify the de-waxing procedure. The proposed method includes construction of a paraffin model based on PCA, EMSC normalization and application of two techniques for spectral quality control. We discuss every step in which the researcher needs to take a subjective decision during the de-waxing procedure, and we explain how to make an adequate choice of parameters involved. Application of this procedure to 71 hyperspectral images collected from 55 human colon biopsies (20 normal, 17 ulcerative colitis, and 18 adenocarcinoma) showed that paraffin was appropriately neutralized, which made the de-waxed images adequate for analysis by pattern-recognition techniques such as k-means clustering or PCA-LDA.

Development of a memetic clustering algorithm for optimal spectral histology: application to FTIR images of normal human colon

The coupling between Fourier-transform infrared (FTIR) imaging and unsupervised classification is effective in revealing the different structures of human tissues based on their specific biomolecular IR signatures; thus the spectral histology of the studied samples is achieved. However, the most widely applied clustering methods in spectral histology are local search algorithms, which converge to a local optimum, depending on initialization. Multiple runs of the techniques estimate multiple different solutions. Here, we propose a memetic algorithm, based on a genetic algorithm and a k-means clustering refinement, to perform optimal clustering. In addition, this approach was applied to the acquired FTIR images of normal human colon tissues originating from five patients. The results show the efficiency of the proposed memetic algorithm to achieve the optimal spectral histology of these samples, contrary to k-means.

Label-free molecular imaging of the kidney

In this review, we will highlight technologies that enable scientists to study the molecular characteristics of tissues and/or cells without the need for antibodies or other labeling techniques. Specifically, we will focus on matrix-assisted laser desorption/ionization imaging mass spectrometry, infrared spectroscopy, and Raman spectroscopy.

Fully unsupervised inter-individual IR spectral histology of paraffinized tissue sections of normal colon

In label-free Fourier-transform infrared histology, spectral images are individually recorded from tissue sections, pre-processed and clustered. Each single resulting color-coded image is annotated by a pathologist to obtain the best possible match with tissue structures revealed after Hematoxylin-Eosin staining. However, the main limitations of this approach are the empirical choice of the number of clusters in unsupervised classification, and the marked color heterogeneity between the clustered spectral images. Here, using normal murine and human colon tissues, we developed an automatic multi-image spectral histology to simultaneously analyze a set of spectral images (8 images mice samples and 72 images human ones). This procedure consisted of a joint Extended Multiplicative Signal Correction (EMSC) to numerically deparaffinize the tissue sections, followed by an automated joint K-Means (KM) clustering using the hierarchical double application of Pakhira-Bandyopadhyay-Maulik (PBM) validity index. Using this procedure, the main murine and human colon histological structures were correctly identified at both the intra- and the inter-individual levels, especially the crypts, secreted mucus, lamina propria and submucosa. Here, we show that batched multi-image spectral histology procedure is insensitive to the reference spectrum but highly sensitive to the paraffin model of joint EMSC. In conclusion, combining joint EMSC and joint KM clustering by double PBM application allows to achieve objective and automated batched multi-image spectral histology.