Dataset Information

Detection of Phenotypic Alterations Using High-Content Analysis of Whole-Slide Images.

ABSTRACT: Tumors exhibit spatial heterogeneity, as manifested in immunohistochemistry (IHC) staining patterns. Current IHC quantification methods lose information by reducing this heterogeneity in each whole-slide image (WSI) or in selective fields of view to a single staining index. The aim of this study was to investigate the sensitivity of an IHC quantification method that uses this heterogeneity to reliably compare IHC staining patterns. We virtually partitioned WSIs by a grid of square tiles, and computed the staining index distributions to quantify heterogeneities. We used samples from these distributions as inputs to non-parametric statistical comparisons. We applied our grid method to fixed tumor samples from 26 tumors obtained from a double-blind preclinical study of a patient-derived orthotopic xenograft model of pediatric neuroblastoma in CD1 nude mice. We compared the results of our grid method to the results based on whole-slide indices, the current practice. We show that our grid method reliably detects phenotypic alterations that other tests based on whole-slide indices fail to detect. Based on robustness and increased sensitivity of statistical inference, we conclude that our method of whole-slide grid quantification is superior to existing whole-slide quantification techniques.

SUBMITTER: Shirinifard A

PROVIDER: S-EPMC4851271 | biostudies-literature | 2016 May

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Detection of Phenotypic Alterations Using High-Content Analysis of Whole-Slide Images.

Shirinifard Abbas A Thiagarajan Suresh S Vogel Peter P Sablauer András A

The journal of histochemistry and cytochemistry : official journal of the Histochemistry Society 20160329 5

Tumors exhibit spatial heterogeneity, as manifested in immunohistochemistry (IHC) staining patterns. Current IHC quantification methods lose information by reducing this heterogeneity in each whole-slide image (WSI) or in selective fields of view to a single staining index. The aim of this study was to investigate the sensitivity of an IHC quantification method that uses this heterogeneity to reliably compare IHC staining patterns. We virtually partitioned WSIs by a grid of square tiles, and com ...[more]

PMID: 27026297

Similar Datasets

Project description:Cervical intraepithelial neoplasia (CIN) and cervical cancer are major health problems faced by women worldwide. The conventional Papanicolaou (Pap) smear analysis is an effective method to diagnose cervical pre-malignant and malignant conditions by analyzing swab images. Various computer vision techniques can be explored to identify potential precancerous and cancerous lesions by analyzing the Pap smear image. The majority of existing work cover binary classification approaches using various classifiers and Convolution Neural Networks. However, they suffer from inherent challenges for minute feature extraction and precise classification. We propose a novel methodology to carry out the multiclass classification of cervical cells from Whole Slide Images (WSI) with optimum feature extraction. The actualization of Conv Net with Transfer Learning technique substantiates meaningful Metamorphic Diagnosis of neoplastic and pre-neoplastic lesions. As the Progressive Resizing technique (an advanced method for training ConvNet) incorporates prior knowledge of the feature hierarchy and can reuse old computations while learning new ones, the model can carry forward the extracted morphological cell features to subsequent Neural Network layers iteratively for elusive learning. The Progressive Resizing technique superimposition in consultation with the Transfer Learning technique while training the Conv Net models has shown a substantial performance increase. The proposed binary and multiclass classification methodology succored in achieving benchmark scores on the Herlev Dataset. We achieved singular multiclass classification scores for WSI images of the SIPaKMed dataset, that is, accuracy (99.70%), precision (99.70%), recall (99.72%), F-Beta (99.63%), and Kappa scores (99.31%), which supersede the scores obtained through principal methodologies. GradCam based feature interpretation extends enhanced assimilation of the generated results, highlighting the pre-malignant and malignant lesions by visual localization in the images.

Project description:BackgroundDeep learning has revolutionized digital pathology, allowing automatic analysis of hematoxylin and eosin (H&E) stained whole slide images (WSIs) for diverse tasks. WSIs are broken into smaller images called tiles, and a neural network encodes each tile. Many recent works use supervised attention-based models to aggregate tile-level features into a slide-level representation, which is then used for downstream analysis. Training supervised attention-based models is computationally intensive, architecture optimization of the attention module is non-trivial, and labeled data are not always available. Therefore, we developed an unsupervised and fast approach called SAMPLER to generate slide-level representations.MethodsSlide-level representations of SAMPLER are generated by encoding the cumulative distribution functions of multiscale tile-level features. To assess effectiveness of SAMPLER, slide-level representations of breast carcinoma (BRCA), non-small cell lung carcinoma (NSCLC), and renal cell carcinoma (RCC) WSIs of The Cancer Genome Atlas (TCGA) were used to train separate classifiers distinguishing tumor subtypes in FFPE and frozen WSIs. In addition, BRCA and NSCLC classifiers were externally validated on frozen WSIs. Moreover, SAMPLER's attention maps identify regions of interest, which were evaluated by a pathologist. To determine time efficiency of SAMPLER, we compared runtime of SAMPLER with two attention-based models. SAMPLER concepts were used to improve the design of a context-aware multi-head attention model (context-MHA).FindingsSAMPLER-based classifiers were comparable to state-of-the-art attention deep learning models to distinguish subtypes of BRCA (AUC = 0.911 ± 0.029), NSCLC (AUC = 0.940 ± 0.018), and RCC (AUC = 0.987 ± 0.006) on FFPE WSIs (internal test sets). However, training SAMLER-based classifiers was >100 times faster. SAMPLER models successfully distinguished tumor subtypes on both internal and external test sets of frozen WSIs. Histopathological review confirmed that SAMPLER-identified high attention tiles contained subtype-specific morphological features. The improved context-MHA distinguished subtypes of BRCA and RCC (BRCA-AUC = 0.921 ± 0.027, RCC-AUC = 0.988 ± 0.010) with increased accuracy on internal test FFPE WSIs.InterpretationOur unsupervised statistical approach is fast and effective for analyzing WSIs, with greatly improved scalability over attention-based deep learning methods. The high accuracy of SAMPLER-based classifiers and interpretable attention maps suggest that SAMPLER successfully encodes the distinct morphologies within WSIs and will be applicable to general histology image analysis problems.FundingThis study was supported by the National Cancer Institute (Grant No. R01CA230031 and P30CA034196).

Dataset Information

Detection of Phenotypic Alterations Using High-Content Analysis of Whole-Slide Images.

Publications

Detection of Phenotypic Alterations Using High-Content Analysis of Whole-Slide Images.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets