Weis, C.-A. | Search bioRxiv

bioRxiv2024

Unsupervised learning for labeling global glomerulosclerosis

Current deep learning models for classifying glomeruli in nephropathology are trained almost exclusively in a supervised manner, requiring expert-labeled images. Very little is known about the potential for unsupervised learning to overcome this bottleneck. To address this open question in a proof-of-concept, the project focused on the most fundamental classification task: globally sclerosed versus non-globally sclerosed glomeruli. The performance of clustering between the two classes was extensively studied across a variety of labeled datasets with diverse compositions and histological stains, and across the feature embeddings produced by 34 different pre-trained CNN models. As demonstrated by the study, clustering of globally and non-globally sclerosed glomeruli is generally highly feasible, yielding accuracies of over 95% in most datasets. Further work will be required to expand these experiments towards the clustering of additional glomerular lesion categories. We are convinced that these efforts (i) will open up opportunities for semi-automatic labeling approaches, thus alleviating the need for labor-intensive manual labeling, and (ii) illustrate that glomerular classification models can potentially be trained even in the absence of expert-derived class labels.

pathology↗

bioRxiv2023

Spatial omics imaging of fresh-frozen tissue and routine FFPE histopathology on a single cancer needle core biopsy: freezing device and multimodal workflow

Complex molecular alterations underlying cancer pathophysiology are intensely studied with omics methods using bulk tissue extracts. For spatially resolved tissue diagnostics using needle biopsy cores, however, histopathological analysis using stained FFPE tissue and immuno-histochemistry (IHC) of few marker proteins is currently the main clinical focus. Today, spatial omics imaging using MSI or IRI are emerging diagnostic technologies for identification and classification of various cancer types. However, to conserve tissue-specific metabolomic states, fast, reliable and precise methods for preparation of fresh-frozen (FF) tissue sections are crucial. Such methods are often incompatible with clinical practice, since spatial metabolomics and routine histopathology of needle biopsies currently require two biopsies for FF and FFPE sampling, respectively. Therefore, we developed a device and corresponding laboratory and computational workflows for multimodal spatial omics analysis of fresh-frozen, longitudinally sectioned needle biopsies to accompany standard FFPE histopathology on the same biopsy core. As proof-of-concept, we analyzed surgical human liver cancer specimen by IRI and MSI with precise co-registration and, following FFPE processing, by sequential clinical pathology analysis on the same biopsy core. This workflow allowed spatial comparison between different spectral profiles and alterations in tissue histology, as well as direct comparison to histological diagnosis without the need of an extra biopsy. SIMPLE SUMMARYRoutine clinical approaches for cancer diagnosis demand fast, cost-efficient, and reliable methods, and their implementation within clinical settings. Currently, histopathology is the golden standard for tissue-based clinical diagnosis. Recently, spatially resolved molecular profiling techniques like mass spectrometry imaging (MSI) or infrared spectroscopy imaging (IRI) have increasingly contributed to clinical research, e.g., by differentiation of cancer subtypes using molecular fingerprints. However, adoption of the corresponding workflows in clinical routine remains challenging, especially for fresh-frozen tissue specimen. Here, we present a novel device based on 3D-printing technology, which facilitates sample preparation of needle biopsies for correlated clinical tissue analysis. It enables combination of MSI and IRI on fresh-frozen clinical samples with histopathological examination of the same needle core after formalin-fixation and paraffin-embedding (FFPE). This device and workflow can pave the way for a more profound understanding of biomolecular processes in cancer and, thus, aid more accurate diagnosis. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=180 SRC="FIGDIR/small/528125v1_ufig1.gif" ALT="Figure 1"> View larger version (55K): org.highwire.dtl.DTLVardef@1da99aborg.highwire.dtl.DTLVardef@9ee706org.highwire.dtl.DTLVardef@5153daorg.highwire.dtl.DTLVardef@1581594_HPS_FORMAT_FIGEXP M_FIG C_FIG

cancer biology↗