Probabilistic spatial analysis in quantitative microscopy with uncertainty-aware cell detection using deep Bayesian regression

Gomariz, Álvaro; Portenier, Tiziano; Nombela‐Arrieta, César; Göksel, Orçun

doi:10.1126/sciadv.abi8295

Cited by 5 publications

(3 citation statements)

References 69 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Uncertainty metrics: Following the work of [84], we evaluated the classification uncertainties in terms of Brier score s B (lower is better) and the NLL s NLL (lower is better) over a set of N test samples, defined as,…”

Section: B Implementation and Evaluationmentioning

confidence: 99%

Weakly Supervised Joint Whole-Slide Segmentation and Classification in Prostate Cancer

Pati¹,

Jaume²,

Zeineb³

et al. 2023

Preprint

View full text Add to dashboard Cite

The segmentation and automatic identification of histological regions of diagnostic interest offer a valuable aid to pathologists. However, segmentation methods are hampered by the difficulty of obtaining pixel-level annotations, which are tedious and expensive to obtain for Whole-Slide images (WSI). To remedy this, weakly supervised methods have been developed to exploit the annotations directly available at the image level. However, to our knowledge, none of these techniques is adapted to deal with WSIs. In this paper, we propose WHOLESIGHT, a weakly-supervised method, to simultaneously segment and classify WSIs of arbitrary shapes and sizes. Formally, WHOLESIGHT first constructs a tissue-graph representation of the WSI, where the nodes and edges depict tissue regions and their interactions, respectively. During training, a graph classification head classifies the WSI and produces node-level pseudo labels via post-hoc feature attribution. These pseudo labels are then used to train a node classification head for WSI segmentation. During testing, both heads simultaneously render class prediction and segmentation for an input WSI. We evaluated WHOLESIGHT on three public prostate cancer WSI datasets. Our method achieved state-of-the-art weakly-supervised segmentation performance on all datasets while resulting in better or comparable classification with respect to state-of-the-art weakly-supervised WSI classification methods. Additionally, we quantify the generalization capability of our method in terms of segmentation and classification performance, uncertainty estimation, and model calibration.

show abstract

Section: B Implementation and Evaluationmentioning

confidence: 99%

Weakly Supervised Joint Whole-Slide Segmentation and Classification in Prostate Cancer

Pati¹,

Jaume²,

Zeineb³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Image shows segmentation of Lyve1+ vessel outlines. Details of the performance metrics of the vessel segmentation models can be found in 53,70 . ( C ) Examples of annotated HPCs and HSCs in one 3D representative image of a FL from a Ctnnal1 GFP/wt Hlf tdTom/wt E13.5 embryo…”

Section: Supplementary Figure Legendsmentioning

confidence: 99%

Tissue-scale dynamic mapping of Hematopoietic Stem Cells and supportive niche cells in the fetal liver

Helbling,

Vijaykumar,

Gomariz

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

SummaryThe fetal liver (FL) plays a fundamental role in the ontogeny of the hematopoietic system, by transiently providing a fertile microenvironment for the maturation, proliferation and expansion of fetal hematopoietic progenitors, as well as definitive hematopoietic stem cells (HSCs). Nonetheless, the cellular make up and identity of HSC niches in the FL remain poorly described. Here we employed a combination of 3D quantitative microscopy, bulk mRNA-seq, flow cytometry and cell-specific reporter models, to dissect the spatiotemporal dynamics of putative niche cells and HSCs in the FL microenvironment. We find that at peak stages of FL hematopoiesis, pro-hematopoietic cytokines are promiscuously expressed by endothelial, mesenchymal cells and most prominently by hepatoblasts. These multicellular consortia of parenchymal/stromal cells are spatially predominant in the FL, thus providing unrestricted access to supportive factors throughout the entire tissue. Accordingly, in these early phases HSCs are found scattered across the parenchyma not displaying obvious spatial biases within the broad microanatomy of the FL, but exhibiting clustering behaviors. This highly conducive microenvironment is transient and gets rapidly remodeled through hepatoblast differentiation and sterile inflammatory signaling, leading to the downregulation of hematopoietic factors and the contraction of supportive niches, which temporarily coincide with the exit of HSCs towards emergent BM tissues.

show abstract

“…Some in-depth studies have explored the uncertainty estimation of deep learning in biological scenarios. Using deep Bayesian learning, Gomariz et al proposed a deep learning-based cell detection framework that could output the desired probabilistic predictions [158], where Bayesian regression techniques were used in uncertainty-aware density maps. In this study, MC dropout is used to capture aleatoric and epistemic uncertainty in the training data, which are used to generate spatial epistemic and aleatoric uncertainty maps as additional inputs for the classifier.…”

Section: Transferring Knowledge From Other Large-scalementioning

confidence: 99%

Deep Learning in Cell Image Analysis

Zhou

Deng

et al. 2022

Intell Comput

View full text Add to dashboard Cite

Cell images, which have been widely used in biomedical research and drug discovery, contain a great deal of valuable information that encodes how cells respond to external stimuli and intentional perturbations. Meanwhile, to discover rarer phenotypes, cell imaging is frequently performed in a high-content manner. Consequently, the manual interpretation of cell images becomes extremely inefficient. Fortunately, with the advancement of deep-learning technologies, an increasing number of deep learning-based algorithms have been developed to automate and streamline this process. In this study, we present an in-depth survey of the three most critical tasks in cell image analysis: segmentation, tracking, and classification. Despite the impressive score, the challenge still remains: most of the algorithms only verify the performance in their customized settings, causing a performance gap between academic research and practical application. Thus, we also review more advanced machine learning technologies, aiming to make deep learning-based methods more useful and eventually promote the application of deep-learning algorithms.

show abstract

Probabilistic spatial analysis in quantitative microscopy with uncertainty-aware cell detection using deep Bayesian regression

Cited by 5 publications

References 69 publications

Weakly Supervised Joint Whole-Slide Segmentation and Classification in Prostate Cancer

Weakly Supervised Joint Whole-Slide Segmentation and Classification in Prostate Cancer

Tissue-scale dynamic mapping of Hematopoietic Stem Cells and supportive niche cells in the fetal liver

Deep Learning in Cell Image Analysis

Contact Info

Product

Resources

About