AI-based pathology predicts origins for cancers of unknown primary

Lu, Ming; Chen, Tiffany; Williamson, Drew F. K.; Zhao, Melissa; Shady, Maha; Lipková, Jana; Mahmood, Faisal

doi:10.1038/s41586-021-03512-4

Cited by 422 publications

(296 citation statements)

References 33 publications

Supporting

Mentioning

293

Contrasting

Order By: Relevance

“…Ex vivo tumor assessment for HRD could incorporate orthogonal testing, including functional assessment, interrogation of TME immune status and NGS. An alternative strategy for multi-feature assessment could include application of deep learning and artificial intelligence to histology specimens in order to identify tumors displaying HRD, as has been used to determine tissue of origin of cancers of unknown primary [152,153]. These techniques show promise for identifying germline BRCA mutant breast cancer based on histological features [154].…”

Section: Current Limitations and Next-generation Strategies In Testing For Hrdmentioning

confidence: 99%

Determinants of Homologous Recombination Deficiency in Pancreatic Cancer

Wattenberg

Reiss

2021

Cancers

View full text Add to dashboard Cite

Pancreatic cancer is a treatment-resistant malignancy associated with high mortality. However, defective homologous recombination (HR), a DNA repair mechanism required for high-fidelity repair of double-strand DNA breaks, is a therapeutic vulnerability. Consistent with this, a subset of patients with pancreatic cancer show unique tumor responsiveness to HR-dependent DNA damage triggered by certain treatments (platinum chemotherapy and PARP inhibitors). While pathogenic mutations in HR genes are a major driver of this sensitivity, another layer of diverse tumor intrinsic and extrinsic factors regulate the HR deficiency (HRD) phenotype. Defining the mechanisms that drive HRD may guide the development of novel strategies and therapeutics to induce treatment sensitivity in non-HRD tumors. Here, we discuss the complexity underlying HRD in pancreatic cancer and highlight implications for identifying and treating this distinct subset of patients.

show abstract

Section: Current Limitations and Next-generation Strategies In Testing For Hrdmentioning

confidence: 99%

Determinants of Homologous Recombination Deficiency in Pancreatic Cancer

Wattenberg

Reiss

2021

Cancers

View full text Add to dashboard Cite

show abstract

“…benchmarking) of different technologies. While the earliest studies in 2018 employed a weakly-supervised approach based on a convolutional neural network (CNN) and spatial averaging [10], recent studies have proposed conceptually new technologies, including attention-based methods [11] and multiple-instance learning [1,2,12]. In addition, computational pathology is an applied field which follows trends in basic computer vision research.…”

Section: Introductionmentioning

confidence: 99%

Benchmarking artificial intelligence methods for end-to-end computational pathology

Laleh

Muti

Loeffler

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Artificial intelligence (AI) can extract subtle visual information from digitized histopathology slides and yield scientific insight on genotype-phenotype interactions as well as clinically actionable recommendations. Classical weakly supervised pipelines use an end-to-end approach with residual neural networks (ResNets), modern convolutional neural networks such as EfficientNet, or non-convolutional architectures such as vision transformers (ViT). In addition, multiple-instance learning (MIL) and clustering-constrained attention MIL (CLAM) are being used for pathology image analysis. However, it is unclear how these different approaches perform relative to each other. Here, we implement and systematically compare all five methods in six clinically relevant end-to-end prediction tasks using data from N=4848 patients with rigorous external validation. We show that histological tumor subtyping of renal cell carcinoma is an easy task which approaches successfully solved with an area under the receiver operating curve (AUROC) of above 0.9 without any significant differences between approaches. In contrast, we report significant performance differences for mutation prediction in colorectal, gastric and bladder cancer. Weakly supervised ResNet- and ViT-based workflows significantly outperformed other methods, in particular MIL and CLAM for mutation prediction. As a reason for this higher performance we identify the ability of ResNet and ViT to assign high prediction scores to highly informative image regions with plausible histopathological image features. We make all source codes publicly available at https://github.com/KatherLab/HIA, allowing easy application of all methods on any end-to-end problem in computational pathology.

show abstract

“…Unsurprisingly, CUP poses challenges to determine the appropriate treatments and clinical care. To address this challenge, Lu et al (2021) developed a deep learning-based algorithm known as Tumor Origin Assessment via Deep Learning (TOAD), with the goal to predict the tissue of origin of the primary tumor using routinely acquired histology images. Histology slides from patients were automatically segmented and divided into thousands of small image patches and fed into a convolutional neural network (CNN) with fixed pretrained parameters.…”

Section: Image-level Analysis Methodsmentioning

confidence: 99%

Deep Learning of Histopathology Images at the Single Cell Level

Lee

Lockhart

Xie

et al. 2021

Front. Artif. Intell.

View full text Add to dashboard Cite

The tumor immune microenvironment (TIME) encompasses many heterogeneous cell types that engage in extensive crosstalk among the cancer, immune, and stromal components. The spatial organization of these different cell types in TIME could be used as biomarkers for predicting drug responses, prognosis and metastasis. Recently, deep learning approaches have been widely used for digital histopathology images for cancer diagnoses and prognoses. Furthermore, some recent approaches have attempted to integrate spatial and molecular omics data to better characterize the TIME. In this review we focus on machine learning-based digital histopathology image analysis methods for characterizing tumor ecosystem. In this review, we will consider three different scales of histopathological analyses that machine learning can operate within: whole slide image (WSI)-level, region of interest (ROI)-level, and cell-level. We will systematically review the various machine learning methods in these three scales with a focus on cell-level analysis. We will provide a perspective of workflow on generating cell-level training data sets using immunohistochemistry markers to “weakly-label” the cell types. We will describe some common steps in the workflow of preparing the data, as well as some limitations of this approach. Finally, we will discuss future opportunities of integrating molecular omics data with digital histopathology images for characterizing tumor ecosystem.

show abstract

AI-based pathology predicts origins for cancers of unknown primary

Cited by 422 publications

References 33 publications

Determinants of Homologous Recombination Deficiency in Pancreatic Cancer

Determinants of Homologous Recombination Deficiency in Pancreatic Cancer

Benchmarking artificial intelligence methods for end-to-end computational pathology

Deep Learning of Histopathology Images at the Single Cell Level

Contact Info

Product

Resources

About