Successful Identification of Nasopharyngeal Carcinoma in Nasopharyngeal Biopsies Using Deep Learning

Abstract

Pathologic diagnosis of nasopharyngeal carcinoma (NPC) can be challenging since most cases are nonkeratinizing carcinoma with little differentiation and many admixed lymphocytes. Our aim was to evaluate the possibility to identify NPC in nasopharyngeal biopsies using deep learning. A total of 726 nasopharyngeal biopsies were included. Among them, 100 cases were randomly selected as the testing set, 20 cases as the validation set, and all other 606 cases as the training set. All three datasets had equal numbers of NPC cases and benign cases. Manual annotation was performed. Cropped square image patches of 256 × 256 pixels were used for patch-level training, validation, and testing. The final patch-level algorithm effectively identified NPC patches, with an area under the receiver operator characteristic curve (AUC) of 0.9900. Using gradient-weighted class activation mapping, we demonstrated that the identification of NPC patches was based on morphologic features of tumor cells. At the second stage, whole-slide images were sequentially cropped into patches, inferred with the patch-level algorithm, and reconstructed into images with a smaller size for training, validation, and testing. Finally, the AUC was 0.9848 for slide-level identification of NPC. Our result shows for the first time that deep learning algorithms can identify NPC.

An annotation-free whole-slide training approach to pathological classification of lung cancer types using deep learning

“…Training cancer classifiers without detailed annotations alleviates the burden of experts and allows DNN models to benefit from numerous WSIs with readily available sign-out diagnoses. In previous studies, methods for detecting cancer using strongly supervised models trained with patch-wise annotations 11 – 19 still outperformed weakly supervised models. However, research on weakly supervised models for cancer detection is gaining popularity because annotation is too costly and because models trained through strong supervision are limited by how targets are annotated.…”

“…Restricted by computing limitations, most histopathology studies have used a two-stage patch-based workflow: a patch-level CNN is trained using patches cropped from a WSI, followed by a slide-level algorithm being trained on features extracted by the patch-level model to reveal the final diagnosis. These patch-based methods have yielded successful results in cancer identification 11 – 19 , cancer type classification 14 , 20 , cancer metastasis detection 13 , 16 – 18 , and prognosis analysis 21 , 22 . However, such methods require experienced pathologists to perform substantial annotation.…”

An Annotation-free Whole-slide Training Approach to Pathological Classification of Lung Cancer Types by Deep Neural Network

“…Training cancer classifiers without detailed annotations alleviates the burden of annotation from experts, and allows deep neural network models to benefit from a huge amount of raw slide data with clinical diagnoses. Compared to previous works, methods on detecting cancers with strong supervision models by patch-wise annotations [11][12][13][14][15][16][17][18][19] still outperformed weak supervision models. However, research on weak supervision models for cancer detecting is gradually increasing since the model trained in a strong supervision manner is limited by how targets are annotated.…”

“…Restricted by computing limitations, most histopathology studies used a two-stage patch-based workflow: a patch-level CNNs training using patches cropped from WSI, followed by a slide-level algorithm trained on features extracted by the patch-level model, to learn the final diagnosis. These patch-based methods have yielded successful results such as cancer identification [11][12][13][14][15][16][17][18][19] , cancer types classification 14,20 , cancer metastasis 13,[16][17][18] and prognosis analysis [21][22] .…”

Prediction of 5-year Progression-Free Survival in Advanced Nasopharyngeal Carcinoma with Pretreatment PET/CT using Multi-Modality Deep Learning-based Radiomics

“…However, from our review, there is no study that attempted to develop end-to-end DLR models for predicting the prognoses of NPC patients from PET/CT images. In addition, existing DLR studies [27][28][29][30][32][33][34] are further limited by: (1) they were mainly designed for single imaging modality such as MRI and CT, so their DLR models cannot derive complementary features from multi-modality PET/CT images; and (2) they had limited comparison to the CR methods (e.g., only a few CR methods were chosen for comparison), which undermines the reliability of their conclusions.…”