Histological Image Processing Features Induce a Quantitative Characterization of Chronic Tumor Hypoxia

Automated mesenchymal stem cell segmentation and machine learning-based phenotype classification using morphometric and textural analysis

Purpose: Mesenchymal stem cells (MSCs) have demonstrated clinically relevant therapeutic effects for treatment of trauma and chronic diseases. The proliferative potential, immunomodulatory characteristics, and multipotentiality of MSCs in monolayer culture is reflected by their morphological phenotype. Standard techniques to evaluate culture viability are subjective, destructive, or time-consuming. We present an image analysis approach to objectively determine morphological phenotype of MSCs for prediction of culture efficacy. Approach: The algorithm was trained using phase-contrast micrographs acquired during the early and mid-logarithmic stages of MSC expansion. Cell regions are localized using edge detection, thresholding, and morphological operations, followed by cell marker identification using H-minima transform within each region to differentiate individual cells from cell clusters. Clusters are segmented using marker-controlled watershed to obtain single cells. Morphometric and textural features are extracted to classify cells based on phenotype using machine learning. Results: Algorithm performance was validated using an independent test dataset of 186 MSCs in 36 culture images. Results show 88% sensitivity and 86% precision for overall cell detection and a mean Sorensen-Dice coefficient of 0.849 ± 0.106 for segmentation per image. The algorithm exhibited an area under the curve of 0.816 (CI 95 = 0.769 to 0.886) and 0.787 (CI 95 = 0.716 to 0.851) for classifying MSCs according to their phenotype at early and mid-logarithmic expansion, respectively. Conclusions: The proposed method shows potential to segment and classify low and moderately dense MSCs based on phenotype with high accuracy and robustness. It enables quantifiable and consistent morphology-based quality assessment for various culture protocols to facilitate cytotherapy development.

MRI Distinguishes Tumor Hypoxia Levels of Different Prognostic and Biological Significance in Cervical Cancer

Tumor hypoxia levels range from mild to severe and have different biological and therapeutical consequences but are not easily assessable in patients. Here we present a method based on diagnostic dynamic contrast enhanced (DCE) MRI that reflects a continuous range of hypoxia levels in patients with tumors of cervical cancer. Hypoxia images were generated using an established approach based on pixel-wise combination of DCE-MRI parameters ν _e and K ^trans, representing oxygen consumption and supply, respectively. Using two tumor models, an algorithm to retrieve surrogate measures of hypoxia levels from the images was developed and validated by comparing the MRI-defined levels with hypoxia levels reflected in pimonidazole-stained histologic sections. An additional indicator of hypoxia levels in patient tumors was established on the basis of expression of nine hypoxia-responsive genes; a strong correlation was found between these indicator values and MRI-defined hypoxia levels in 63 patients. Chemoradiotherapy outcome of 74 patients was most strongly predicted by moderate hypoxia levels, whereas more severe or milder levels were less predictive. By combining gene expression profiles and MRI-defined hypoxia levels in cancer hallmark analysis, we identified a distribution of levels associated with each hallmark; oxidative phosphorylation and G₂-M checkpoint were associated with moderate hypoxia, epithelial-to-mesenchymal transition, and inflammatory responses with significantly more severe levels. At the mildest levels, IFN response hallmarks together with HIF1A protein expression by IHC appeared significant. Thus, our method visualizes the distribution of hypoxia levels within patient tumors and has potential to distinguish levels of different prognostic and biological significance. SIGNIFICANCE: These findings present an approach to image a continuous range of hypoxia levels in tumors and demonstrate the combination of imaging with molecular data to better understand the biology behind these different levels.

Learning from irregularly sampled data for endomicroscopy super-resolution: a comparative study of sparse and dense approaches

Purpose

Probe-based confocal laser endomicroscopy (pCLE) enables performing an optical biopsy via a probe. pCLE probes consist of multiple optical fibres arranged in a bundle, which taken together generate signals in an irregularly sampled pattern. Current pCLE reconstruction is based on interpolating irregular signals onto an over-sampled Cartesian grid, using a naive linear interpolation. It was shown that convolutional neural networks (CNNs) could improve pCLE image quality. Yet classical CNNs may be suboptimal in regard to irregular data.

Methods

We compare pCLE reconstruction and super-resolution (SR) methods taking irregularly sampled or reconstructed pCLE images as input. We also propose to embed a Nadaraya–Watson (NW) kernel regression into the CNN framework as a novel trainable CNN layer. We design deep learning architectures allowing for reconstructing high-quality pCLE images directly from the irregularly sampled input data. We created synthetic sparse pCLE images to evaluate our methodology.

Results

The results were validated through an image quality assessment based on a combination of the following metrics: peak signal-to-noise ratio and the structural similarity index. Our analysis indicates that both dense and sparse CNNs outperform the reconstruction method currently used in the clinic.

Conclusion

The main contributions of our study are a comparison of sparse and dense approach in pCLE image reconstruction. We also implement trainable generalised NW kernel regression as a novel sparse approach. We also generated synthetic data for training pCLE SR.

Electronic supplementary material

The online version of this article (10.1007/s11548-020-02170-7) contains supplementary material, which is available to authorized users.

Tumor apelin immunoreactivity is correlated with body mass index in ovarian high grade serous carcinoma

Ovarian cancer has a high mortality rate. Serous carcinoma is the most common subtype and can be detected by distant or lymph node metastasis in advanced stages. Apelin, an adipokine associated with obesity, and its receptor, APJ, participate in lymphatic invasion. Angiogenesis also can affect lymph node involvement in serous ovarian carcinomas. We investigated apelin/APJ receptor immunoreactivity in stages III and IV ovarian cancer with or without lymph node involvement and correlated the results with body mass index (BMI) to determine whether the potential relation of the two affects the outcome of the cancer. We investigated 30 patients diagnosed between 2014 and 2016 with high grade serous ovarian cancer. Tumor:stroma ratio, indirect immunoperoxidase method, H-score and MATLAB analysis were performed. In obese and pre-obese patients, tumor apelin immunoreactivity was stronger than for patients with normal BMI. Tumor:stroma ratio was correlated with survival and lymph node involvement. Strong apelin and moderate APJ immunoreactivity was detected in both lymph node negative and positive patients. BMI was related to both survival outcome and apelin immunoreactivity. BMI, adipokines such as apelin, and the stromal compartment play critical roles in advanced stage serous carcinomas.