. 2023 Dec 18;13(24):3685.
doi: 10.3390/diagnostics13243685.
Affiliations
Free PMC article
Item in Clipboard
Diagnostics (Basel).
.
Free PMC article
Abstract
Nasopharyngeal carcinoma (NPC) is an epithelial cancer originating in the nasopharynx epithelium. Nevertheless, annotating pathology slides remains a bottleneck in the development of AI-driven pathology models and applications. In the present study, we aim to demonstrate the feasibility of using immunohistochemistry (IHC) for annotation by non-pathologists and to develop an efficient model for distinguishing NPC without the time-consuming involvement of pathologists. For this study, we gathered NPC slides from 251 different patients, comprising hematoxylin and eosin (H&E) slides, pan-cytokeratin (Pan-CK) IHC slides, and Epstein-Barr virus-encoded small RNA (EBER) slides. The annotation of NPC regions in the H&E slides was carried out by a non-pathologist trainee who had access to corresponding Pan-CK IHC slides, both with and without EBER slides. The training process utilized ResNeXt, a deep neural network featuring a residual and inception architecture. In the validation set, NPC exhibited an AUC of 0.896, with a sensitivity of 0.919 and a specificity of 0.878. This study represents a significant breakthrough: the successful application of deep convolutional neural networks to identify NPC without the need for expert pathologist annotations. Our results underscore the potential of laboratory techniques to substantially reduce the workload of pathologists.
Keywords:
diagnose; immunohistochemistry; machine learning; nasopharyngeal carcinoma.
Conflict of interest statement
The authors declare no conflict of interest.
Figures
Figure 1
The study design of the current study. NPC, nasopharyngeal carcinoma; Pan-CK, pan-cytokeratin antibody; EBER, Epstein–Barr virus in situ hybridization.
Figure 2
Representative pictures of nasopharyngeal carcinoma and corresponding Pan-CK immunohistochemical stain. Representative pictures of nasopharyngeal carcinoma in H&E stain (A,C) and corresponding Pan-CK immunohistochemical stain (B,D). The easier case with NPC cells (red asterisks) in a large pale morphology in H&E slide (A) and highlighted by Pan-CK IHC stain (B). Harder NPC cases with NPC cells infiltrating (red circle) in the stroma in a small nest showed in the H&E slide (C) and also highlighted by Pan-CK IHC stain (D). All figures are in 100× magnification.
Figure 3
Representative pictures of difficult nasopharyngeal carcinoma case with H&E stain and corresponding Pan-CK immunohistochemical stain and EBER stain. Representative pictures of difficult nasopharyngeal carcinoma cases, with mostly single or tiny tumor nests (red arrows) in H&E stain (A) and corresponding Pan-CK IHC (B) and EBER stain (C). All figures are in 100× magnification.
Figure 4
Representative pictures of free-hand region-of-interest labelling in nasopharyngeal carcinoma slides. The non-pathologist trainee label the NPC area in a free-hand region-of-interest style for NPC areas (green dots area). They labelled the NPC area by simultaneous the area of pan-CK IHC slides with/without EBER slides.
Figure 5
The representative pictures of different size of the patch and model design. The representative NPC cells size in different patch size of 1000 × 1000, 400 × 400, and 200 × 200 (A). In our patch-level training, testing, and validation procedures, we employed square image patches with dimensions of 400 × 400 pixels. These patches were randomly and dynamically cropped from the areas that were free-hand annotated. (B). Our training utilized ResNeXt, a deep neural network architecture incorporating both residual and inception features. During training, we employed a batch size of 48 patches, consisting of 24 patches for NPC and 24 patches for benign tissue.
Figure 6
The training history of the train set and test set. The training history of the accuracy and epochs. After 14 epochs, the accuracy reached more than 85% in both the train and test set.
References
-
-
Ferlay J., Ervik M., Lam F., Colombet M., Mery L., Piñeros M. Global Cancer Observatory: Cancer Today. International Agency for Research on Cancer; Lyon, France: 2018. [(accessed on 22 October 2023)]. Available online: https://gco.iarc.fr/today.
-
Grants and funding
This research received no external funding.
LinkOut – more resources
-
Full Text Sources
-
Research Materials
