Robust subtyping of non‐small cell lung cancer whole sections through MALDI mass spectrometry imaging

Abstract

Subtyping of the most common non‐small cell lung cancer (NSCLC) tumor types adenocarcinoma (ADC) and squamous cell carcinoma (SqCC) is still a challenge in the clinical routine and a correct diagnosis is crucial for an adequate therapy selection. Matrix‐assisted laser desorption/ionization (MALDI) mass spectrometry imaging (MSI) has shown potential for NSCLC subtyping but is subject to strong technical variability and has only been applied to tissue samples assembled in tissue microarrays (TMAs). To our knowledge, a successful transfer of a classifier from TMAs to whole sections, which are generated in the standard clinical routine, has not been presented in the literature as of yet. We introduce a classification algorithm using extensive preprocessing and a classifier (either a neural network or a linear discriminant analysis (LDA)) to robustly classify whole sections of ADC and SqCC lung tissue. The classifiers were trained on TMAs and validated and tested on whole sections. Vital for a successful application on whole sections is the extensive preprocessing and the use of whole sections for hyperparameter selection. The classification system with the neural network/LDA results in 99.0%/98.3% test accuracy on spectra level and 100.0%/100.0% test accuracy on whole section level, respectively, and, therefore, provides a powerful tool to support the pathologist's decision making process. The presented method is a step further towards a clinical application of MALDI MSI and artificial intelligence for subtyping of NSCLC tissue sections.

Multi-Class Cancer Subtyping in Salivary Gland Carcinomas with MALDI Imaging and Deep Learning

“…Regarding the applicability to a clinical setting, a recently published work shows that machine learning algorithms trained on MALDI imaging measurement of TMAs can successfully be transferred to classification tasks on whole sections [ 32 ]. While these results were obtained with lung cancer patients, it seems likely that these results hold true for the tissue types and classification tasks presented in our study, since the experimental setup shows strong similarities and in both studies the task is cancer subtyping.…”

Are mass spectrometry imaging‐based diagnostics becoming reality?

“…In their article in this issue, “Robust Subtyping of Non‐Small Cell Lung Cancer Whole Sections through MALDI Mass Spectrometry Imaging,” [4] Janßen et al. have sought to address the use of TMAs to develop a classification algorithm that is applicable to traditional biopsies as well as standardizing a data processing pipeline for optimal classification results.…”

“…taken from dozens to hundreds of biopsies and collected into a single wax block that can be sectioned and subjected to standard analytical techniques such as histological staining or immunohistochemistry.More recently, TMAs have been used for MSI studies allowing for hundreds of patient samples to be analyzed in a single experiment, greatly minimizing the amount of time needed to generate enough data to perform training and validation studies[2,3]. Given the small size of tissue specimens used for these studies, concerns exist on how well the original biopsies are represented, and if these results will translate to the analysis of new biopsy specimens.In their article in this issue, "Robust Subtyping of Non-Small Cell Lung Cancer Whole Sections through MALDI Mass Spectrometry Imaging,"[4] Janßen et al have sought to address the use of TMAs to develop a classification algorithm that is applicable to traditional biopsies as well as standardizing a data processing pipeline for optimal classification results. Two models (linear discriminant analysis and neural networks) to differentiate between adenocarcinomas and squamous cell carcinomas of the lung were trained using multiple TMAs and then validated and tested on cancerous areas of whole biopsy sections, obtaining 100% test set sample classification accuracy.…”

Multimodal Lung Cancer Subtyping Using Deep Learning Neural Networks on Whole Slide Tissue Images and MALDI MSI

“…Machine learning techniques typically require large data cohorts that show a high biological variation and hence these methods are usually developed and evaluated on tissue micro arrays (TMAs), the construction of which is not feasible in standard clinical routine. Recently, we presented an AI-based classification algorithm for discriminating ADC and SqCC of the lung based on MALDI MSI data from whole tissue sections [ 21 ]. This method, however, still requires a pathologist to annotate tumor areas.…”

“…The network architecture from Le’Clerc Arrastia et al [ 22 ] was used for the segmentation of the WSIs to detect areas with high tumor cell content. Subsequently, the respective co-registered MALDI MSI data were classified into ADC and SqCC using a previously developed neural network [ 20 , 21 ].…”