Deep learning pathological microscopic features in endemic nasopharyngeal cancer: Prognostic value and protentional role for individual induction chemotherapy

Abstract

Background To explore the prognostic value and the role for treatment decision of pathological microscopic features in patients with nasopharyngeal carcinoma (NPC) using the method of deep learning. Methods The pathological microscopic features were extracted using the software QuPath (version 0.1.3. Queen's University) in the training cohort (Guangzhou training cohort, n = 843). We used the neural network DeepSurv to analyze the pathological microscopic features (DSPMF) and then classified patients into high‐risk and low‐risk groups through the time‐dependent receiver operating characteristic (ROC). The prognosis accuracy of the pathological feature was validated in a validation cohort (n = 212). The primary endpoint was progression‐free survival (PFS). Results We found 429 pathological microscopic features in the H&E image. Patients with high‐risk scores in the training cohort had shorter 5‐year PFS (HR 10.03, 6.06‐16.61; P < .0001). The DSPMF (C‐index: 0.723) had the higher C‐index than the EBV DNA (C‐index: 0.612) copies and the N stage (C‐index: 0.593). Furthermore, induction chemotherapy (ICT) plus concomitant chemoradiotherapy (CCRT) had better 5‐year PFS to those received CCRT (P < .0001) in the high‐risk group. Conclusion The DSPMF is a reliable prognostic tool for survival risk in patients with NPC and might be able to guide the treatment decision.

A deep-learning-based prognostic nomogram integrating microscopic digital pathology and macroscopic magnetic resonance images in nasopharyngeal carcinoma: a multi-cohort study

“…Impressively, we found that clinic, histopathologic, and radiomic signatures exhibited complementary value in pretreatment individualized prediction of prognoses, with the best performance in pretreatment individualized prediction of prognoses in two test cohorts (Table 2). Compared with previous radiomic and histopathologic models (C-index 0.723-0.761), 10,19 our results also showed an impressive improvement (C-index > 0.8 in three cohorts). We believed these improvements might come from the maximum mining of clinic data on different spatial scales.…”

“…A recent study about NPC showed a signature based on handcrafted features from WSI as an independent prognostic factor, with a similar performance (C-index 0.723). 19 In contrast, our histopathologic signature was constructed in an end-to-end method, avoiding incompleteness and instabilities of artificial setting. To interpret the histopathologic signature, we visualized DCNN using the attention map, 30 and found the hotspots located on the tumor cells with multiple nucleoli ( Figure 4a ), vesicular nuclei ( Figure 4b ), and spindle shape ( Figure 4c , as sarcomatoid tumor cells).…”

“…To date, the underlying biology of imaging predictors for NPC have barely been studied before. 10 , 12 , 13 , 19 , 31 Herein, we performed a preliminary investigation of the association between gene expression and the imaging features of our model. Notably, we identified two texture features related to the genetic alterations of chromatin remodeling pathway, which is regarded as a mark of subclone diversity during tumor evolution, and is correlated to higher mutational burden, EBV burden, and poor prognoses.…”

“… 14 – 16 Histopathologic properties such as histopathologic classification, 17 immunological phenotype [such as tumor-infiltrating lymphocytes (TILs)] 18 and handcrafted microscopic features from WSI are found to be independently prognostic for NPC. 19 However, there have been no published reports on the integration of imaging features from radiological images and WSI for the prognostic prediction of patients with cancer including NPC. Therefore, it is necessary and reasonable to integrate histopathologic and radiomic features, as well as the known prognostic factors, to assess the morphological features at multiple scale (cell, tissue, and patient level), thus enabling better characterization of the aggressiveness of disease.…”

QuPath: The global impact of an open source digital pathology system

“…In developing the prognostic use of QuPath as a tool for annotation training for deep learning, Liu et al [40] showed in 843 samples that their network could stratify patients with nasopharyngeal carcinoma into a high risk group with shorter than 5 year progression free survival (p < 0.0001).…”