Soft tissue sarcomas (STS) are considered non-immunogenic, although distinct entities respond to antitumor agents targeting the tumor microenvironment. This study's aims were to investigate relationships between tumor-infiltrating immune cells and patient/tumor-related factors, and assess their prognostic value for local recurrence (LR), distant metastasis (DM), and overall survival (OS). One-hundred-eighty-eight STS-patients (87 females [46.3%]; median age: 62.5 years) were retrospectively analyzed. Tissue microarrays (in total 1266 cores) were stained with multiplex immunohistochemistry and analyzed with multispectral imaging. Seven cell types were differentiated depending on marker profiles (CD3+, CD3+ CD4+ helper, CD3+ CD8+ cytotoxic, CD3+ CD4+ CD45RO+ helper memory, CD3 + CD8+ CD45RO+ cytotoxic memory T-cells; CD20 + B-cells; CD68+ macrophages). Correlations between phenotype abundance and variables were analyzed. Uni-and multivariate Fine&Gray and Cox-regression models were constructed to investigate prognostic variables. Model calibration was assessed with C-index. IHC-findings were validated with TCGA-SARC gene expression data of genes specific for macrophages, T-and B-cells. B-cell percentage was lower in patients older than 62.5 years (p = .013), whilst macrophage percentage was higher (p = .002). High B-cell (p = .035) and macrophage levels (p = .003) were associated with increased LR-risk in the univariate analysis. In the multivariate setting, high macrophage levels (p = .014) were associated with increased LR-risk, irrespective of margins, age, gender or B-cells. Other immune cells were not associated with outcome events. High macrophage levels were a poor prognostic factor for LR, irrespective of margins, B-cells, gender and age. Thus, anti-tumor, macrophage-targeting agents may be applied more frequently in tumors with enhanced macrophage infiltration.

No abstract

No abstract

Purpose: Cell segmentation algorithms are commonly used to analyze large histologic images as they facilitate interpretation, but on the other hand they complicate hypothesis-free spatial analysis. Therefore, many applications train convolutional neural networks (CNNs) on highresolution images that resolve individual cells instead, but their practical application is severely limited by computational resources. In this work, we propose and investigate an alternative spatial data representation based on cell segmentation data for direct training of CNNs.Approach: We introduce and analyze the properties of Cell2Grid, an algorithm that generates compact images from cell segmentation data by placing individual cells into a low-resolution grid and resolves possible cell conflicts. For evaluation, we present a case study on colorectal cancer relapse prediction using fluorescent multiplex immunohistochemistry images.Results: We could generate Cell2Grid images at 5-μm resolution that were 100 times smaller than the original ones. Cell features, such as phenotype counts and nearest-neighbor cell distances, remain similar to those of original cell segmentation tables (p < 0.0001). These images could be directly fed to a CNN for predicting colon cancer relapse. Our experiments showed that test set error rate was reduced by 25% compared with CNNs trained on images rescaled to 5μm with bilinear interpolation. Compared with images at 1-μm resolution (bilinear rescaling), our method reduced CNN training time by 85%. Conclusions:Cell2Grid is an efficient spatial data representation algorithm that enables the use of conventional CNNs on cell segmentation data. Its cell-based representation additionally opens a door for simplified model interpretation and synthetic image generation.

In this paper we propose C2G-Net, a pipeline for image classification that exploits the morphological properties of images containing a large number of similar objects like biological cells. C2G-Net consists of two components: (1) Cell2Grid, an image compression algorithm that identifies objects using segmentation and arranges them on a grid, and (2) DeepLNiNo, a CNN architecture with less than 10,000 trainable parameters aimed at facilitating model interpretability. To test the performance of C2G-Net we used multiplex immunohistochemistry images for predicting relapse risk in colon cancer. Compared to conventional CNN architectures trained on raw images, C2G-Net achieved similar prediction accuracy while training time was reduced by 85% and its model was is easier to interpret.