Dataset Information

Deep learning systems detect dysplasia with human-like accuracy using histopathology and probe-based confocal laser endomicroscopy.

ABSTRACT: Probe-based confocal laser endomicroscopy (pCLE) allows for real-time diagnosis of dysplasia and cancer in Barrett's esophagus (BE) but is limited by low sensitivity. Even the gold standard of histopathology is hindered by poor agreement between pathologists. We deployed deep-learning-based image and video analysis in order to improve diagnostic accuracy of pCLE videos and biopsy images. Blinded experts categorized biopsies and pCLE videos as squamous, non-dysplastic BE, or dysplasia/cancer, and deep learning models were trained to classify the data into these three categories. Biopsy classification was conducted using two distinct approaches-a patch-level model and a whole-slide-image-level model. Gradient-weighted class activation maps (Grad-CAMs) were extracted from pCLE and biopsy models in order to determine tissue structures deemed relevant by the models. 1970 pCLE videos, 897,931 biopsy patches, and 387 whole-slide images were used to train, test, and validate the models. In pCLE analysis, models achieved a high sensitivity for dysplasia (71%) and an overall accuracy of 90% for all classes. For biopsies at the patch level, the model achieved a sensitivity of 72% for dysplasia and an overall accuracy of 90%. The whole-slide-image-level model achieved a sensitivity of 90% for dysplasia and 94% overall accuracy. Grad-CAMs for all models showed activation in medically relevant tissue regions. Our deep learning models achieved high diagnostic accuracy for both pCLE-based and histopathologic diagnosis of esophageal dysplasia and its precursors, similar to human accuracy in prior studies. These machine learning approaches may improve accuracy and efficiency of current screening protocols.

SUBMITTER: Guleria S

PROVIDER: S-EPMC7930108 | biostudies-literature | 2021 Mar

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Deep learning systems detect dysplasia with human-like accuracy using histopathology and probe-based confocal laser endomicroscopy.

Guleria Shan S Shah Tilak U TU Pulido J Vincent JV Fasullo Matthew M Ehsan Lubaina L Lippman Robert R Sali Rasoul R Mutha Pritesh P Cheng Lin L Brown Donald E DE Syed Sana S

Scientific reports 20210303 1

Probe-based confocal laser endomicroscopy (pCLE) allows for real-time diagnosis of dysplasia and cancer in Barrett's esophagus (BE) but is limited by low sensitivity. Even the gold standard of histopathology is hindered by poor agreement between pathologists. We deployed deep-learning-based image and video analysis in order to improve diagnostic accuracy of pCLE videos and biopsy images. Blinded experts categorized biopsies and pCLE videos as squamous, non-dysplastic BE, or dysplasia/cancer, and ...[more]

PMID: 33658592

Similar Datasets

Project description:Background and aimsProbe-based confocal laser endomicroscopy (pCLE) and volumetric laser endomicroscopy (VLE) (also known as frequency domain optical coherence tomography) are advanced endoscopic imaging modalities that may be useful in the diagnosis of dysplasia associated with Barrett's esophagus (BE). We performed pCLE examination in ex-vivo EMR specimens and compared the diagnostic performance of using the current VLE scoring index (previously established as OCT-SI) and a novel VLE diagnostic algorithm (VLE-DA) for the detection of dysplasia.MethodsA total of 27 patients with BE enrolled in a surveillance program at a tertiary-care center underwent 50 clinically indicated EMRs that were imaged with VLE and pCLE and classified into neoplastic (N = 34; high-grade dysplasia, intramucosal adenocarcinoma) and nonneoplastic (N = 16; low-grade dysplasia, nondysplastic BE), based on histology. Image datasets (VLE, N = 50; pCLE, N = 50) were rated by 3 gastroenterologists trained in the established diagnostic criteria for each imaging modality as well as a new diagnostic algorithm for VLE derived from a training set that demonstrated association of specific VLE features with neoplasia. Sensitivity, specificity, and diagnostic accuracy were assessed for each imaging modality and diagnostic criteria.ResultsThe sensitivity, specificity, and diagnostic accuracy of pCLE for detection of BE dysplasia was 76% (95% confidence interval [CI], 59-88), 79% (95% CI, 53-92), and 77% (95% CI, 72-82), respectively. The optimal diagnostic performance of OCT-SI showed a sensitivity of 70% (95% CI, 52-84), specificity of 60% (95% CI, 36-79), and diagnostic accuracy of 67%; (95% CI, 58-78). The use of the novel VLE-DA showed a sensitivity of 86% (95% CI, 69-96), specificity of 88% (95% CI, 60-99), and diagnostic accuracy of 87% (95% CI, 86-88). The diagnostic accuracy of using the new VLE-DA criteria was significantly superior to the current OCT-SI (P < .01).ConclusionThe use of a new VLE-DA showed enhanced diagnostic performance for detecting BE dysplasia ex vivo compared with the current OCT-SI. Further validation of this algorithm in vivo is warranted.

Dataset Information

Deep learning systems detect dysplasia with human-like accuracy using histopathology and probe-based confocal laser endomicroscopy.

Publications

Deep learning systems detect dysplasia with human-like accuracy using histopathology and probe-based confocal laser endomicroscopy.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets