Deep learning for automatic diagnosis of gastric dysplasia using whole-slide histopathology images in endoscopic specimens.

BACKGROUND: Distinguishing gastric epithelial regeneration change from dysplasia and histopathological diagnosis of dysplasia is subject to interobserver disagreement in endoscopic specimens. In this study, we developed a method to distinguish gastric epithelial regeneration change from dysplasia and further subclassify dysplasia. Meanwhile, optimized the cross-hospital diagnosis using domain adaption (DA). METHODS: 897 whole slide images (WSIs) of endoscopic specimens from two hospitals were divided into training, internal validation, and external validation cohorts. We developed a deep learning (DL) with DA (DLDA) model to classify gastric dysplasia and epithelial regeneration change into three categories: negative for dysplasia (NFD), low-grade dysplasia (LGD), and high-grade dysplasia (HGD)/intramucosal invasion neoplasia (IMN). The diagnosis based on the DLDA model was compared to 12 pathologists using 100 gastric biopsy cases. RESULTS: In the internal validation cohort, the diagnostic performance measured by the macro-averaged area under the receiver operating characteristic curve (AUC) was 0.97. In the independent external validation cohort, our DLDA models increased macro-averaged AUC from 0.67 to 0.82. In terms of the NFD and HGD cases, our model's diagnostic sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were significantly higher than junior and senior pathologists. Our model's diagnostic sensitivity, NPV, was higher than specialist pathologists. CONCLUSIONS: We demonstrated that our DLDA model could distinguish gastric epithelial regeneration change from dysplasia and further subclassify dysplasia in endoscopic specimens. Meanwhile, achieved significant improvement of diagnosis cross-hospital.

[Application and Progress of Convolutional Neural Network-based Pathological Diagnosis of Gastric Cancer].

The incidence of gastric cancer is the highest among all kinds of malignant tumors in China. Because gastric cancer is very hard to identify in its early stage, the early diagnosis rate of gastric cancer in China is relatively low. At present, the pathological diagnosis of gastric cancer mainly depends on the diagnosis of pathologists. However, the gradual improvement of people's living standards and the growing demand for medical and health care have exacerbated the shortage of medical resources, which has become a even more serious problem. Therefore, there is an urgent need for new technologies to help deal with this challenge. In recent years, with the rapid development of artificial intelligence (AI) and digital pathology, AI-aided pathological diagnosis based on convolutional neural network (CNN) as the core technology is showing promises for improving the diagnostic efficiency of gastric cancer. It is also of great significance for the early diagnosis and treatment of the disease and the reduction of its high incidence and mortality. We herein summarize the application and progress of deep-learning CNN in pathological diagnosis of gastric cancer, as well as the existing problems and prospects of future development.

The association between PD-1 / PD-L1 expression and clinicopathological features in sarcomatoid renal cell carcinoma.

BACKGROUND: Sarcomatoid renal cell carcinoma (sRCC) accounts for about 4%-5% of all kidney cancers. Previous studies showed that PD-1 and PD-L1 expression was higher in sRCC compared to non-sRCC. In the present study, we aimed to investigate PD-1/PD-L1 expression and its association with clinicopathological features in sRCC. METHODS: The study included 59 patients diagnosed with sRCC between January 2012 and January 2022. The expression of PD-1 and PD-L1 in sRCC was detected by immunohistochemical staining, and its correlation with clinicopathological parameters was analyzed by χ2 test and Fisher exact test. Kaplan-Meier curves and log-rank tests were used to describe the overall survival (OS). The prognostic significance of clinicopathological parameters on OS was assessed by Cox proportional hazards regression analysis. RESULTS: Among the 59 cases, the positive expression of PD-1 and PD-L1 was 34 cases (57.6%) and 37 cases (62.7%), respectively. PD-1 expression was not significantly correlated with any parameters. However, PD-L1 expression was significantly correlated with tumor size and pathologic T stage. OS was shorter in the subgroup of patients with PD-L1-positive sRCC compared with the PD-L1-negative subgroup. There was no statistically significant difference in OS between PD-1-positive and negative subgroups. According to our study, the univariate and multivariate analysis indicated that pathological T3 and T4 was an independent risk factor in PD-1-positive sRCC. CONCLUSION: We studied the relationship between PD-1/PD-L1 expression and clinicopathological characteristics in sRCC. The findings may provide valuable implications for clinical prediction.

DNA methylation profiling identifies epigenetic signatures of early gastric cancer.

Research on the DNA methylation status of gastric cancer (GC) has primarily focused on identifying invasive GC to develop biomarkers for diagnostic. However, DNA methylation in noninvasive GC remains unclear. We conducted a comprehensive DNA methylation profiling study of differentiated-type intramucosal GCs (IMCs). Illumina 850K microarrays were utilized to assess the DNA methylation profiles of formalin-fixed paraffin-embedded tissues from eight patients who were Epstein-Barr virus-negative and DNA mismatch repair proficient, including IMCs and paired adjacent nontumor mucosa. Gene expression profiling microarray data from the GEO database were analyzed via bioinformatics to identify candidate methylation genes. The final validation was conducted using quantitative real-time PCR, the TCGA methylation database, and single-sample gene set enrichment analysis (GSEA). Genome-wide DNA methylation profiling revealed a global decrease in methylation in IMCs compared with nontumor tissues. Differential methylation analysis between IMCs and nontumor tissues identified 449 differentially methylated probes, with a majority of sites showing hypomethylation in IMCs compared with nontumor tissues (66.1% vs 33.9%). Integrating two RNA-seq microarray datasets, we found one hypomethylation-upregulated gene: eEF1A2, overlapped with our DNA methylation data. The mRNA expression of eEF1A2 was higher in twenty-four IMC tissues than in their paired adjacent nontumor tissues. GSEA indicated that the functions of eEF1A2 were associated with the development of IMCs. Furthermore, TCGA data indicated that eEF1A2 is hypomethylated in advanced GC. Our study illustrates the implications of DNA methylation alterations in IMCs and suggests that aberrant hypomethylation and high mRNA expression of eEF1A2 might play a role in IMCs development.

Morphological Characteristics of Various Cells in Esophageal Squamous Dysplasia: Extremely Wide Morphological Spectrum.

The WHO classification of esophageal tumors divides esophageal squamous intraepithelial dysplasia into high and low grades, but does not specify its morphological spectrum. Here, the morphological characteristics of various cells were investigated in esophageal squamous (high-grade) dysplasia, and a morphological spectrum and terminology for this lesion were proposed to avoid misdiagnosis. The clinicopathological data of 540 patients with esophageal squamous dysplasia were analyzed retrospectively. According to the unique cytomorphological characteristics of the lesions and the predominant cell type, the esophageal squamous dysplasia was divided into the following morphological groups: classic type (34.6%, 187/540), basaloid subtype (10.7%, 58/540), spindle-cell subtype (4.6%, 25/540), differentiated subtype (48.9%, 264/540), and verrucous subtype (1.1%, 6/540). Gender, age, and lesions location did not differ among the subtypes (P > 0.05), while Paris classification and lesions diameter significantly differed among the subtypes (P < 0.01). Classic-type cells showed severe atypia. In the basaloid subtype, the cells were small, and resembled basal cells; most of these lesions were of the 0-IIb type with small lesion diameter. In the spindle-cell subtype, the cells and nuclei were spindle-shaped or long and spindle-shaped and arranged in parallel. Differentiated-subtype showed well-to-moderately differentiated cells, and epithelial basal cells were mature. Verrucous-subtype showed well-differentiated cells, and were characterized by verrucous or papillary structures. Esophageal squamous dysplasia has extremely wide morphological spectrum. Awareness of the spectrum of morphological presentations of this lesion, specifically the basaloid subtype, spindle-cell subtype, differentiated subtype, and verrucous subtype, is important for accurate diagnosis.

Reticular fibre structure in the differential diagnosis of parathyroid neoplasms.

BACKGROUND: To investigate the characteristics of reticular fibre structure (RFS) in parathyroid adenoma (PTA), atypical parathyroid tumour (APT), and parathyroid carcinoma (PTC), and to assess its value as a diagnostic indicator. METHODS: Clinical data and pathological specimens of patients with PTA, APT or PTC were collected. Reticular fibre staining was performed to observe the characteristics of RFS. This study evaluated the incidence of RFS destruction in parathyroid tumours, compared RFS destruction between primary PTC and recurrent and metastatic PTC, and explored the association between RFS destruction and clinicopathological features of APT and primary PTC. RESULTS: Reticular fibre staining was performed in 50 patients with PTA, 25 patients with APT, and 36 patients with PTC. In PTA cases, a delicate RFS was observed. In both the APT and PTC groups, incomplete RFS areas were observed. The incidence of RFS destruction was different among the PTA, APT, and PTC groups (P < 0.001, χ2-test), at 0% (0/50), 44% (11/25), and 86% (31/36), respectively. When differentiating PTC from APT, the sensitivity and specificity of RFS destruction were 81% and 56%, respectively. The incidence of RFS destruction was 73% (8/11) in the primary PTC group and 92% (23/25) in the recurrent and metastatic PTC groups. In both the APT group and primary PTC group, no correlation was found between RFS destruction and clinicopathological features. CONCLUSION: RFS destruction may indicate that parathyroid tumours have unfavourable biological behaviours.Reticular fibre staining may be a valuable tool for improving the diagnostic accuracy in parathyroid tumours.

Deep learning-assisted development and validation of an algorithm for predicting the growth of persistent pure ground-glass nodules.

Background: The prediction of the persistent pure ground-glass nodule (pGGN) growth is challenging and limited by subjective assessment and variation across radiologists. A chest computed tomography (CT) image-based deep learning classification model (DLCM) may provide a more accurate growth prediction. Methods: This retrospective study enrolled consecutive patients with pGGNs from January 2010 to December 2020 from two independent medical institutions. Four DLCM algorithms were built to predict the growth of pGGNs, which were extracted from the nodule areas of chest CT images annotated by two radiologists. All nodules were assigned to either the study, the inner validation, or the external validation cohort. Accuracy, sensitivity, specificity, receiver operating characteristic (ROC) curves, and areas under the ROC curve (AUROCs) were analyzed to evaluate our models. Results: A total of 286 patients were included, with 419 pGGN. In total, 197 (68.9%) of the patients were female and the average age was 59.5±12.0 years. The number of pGGN assigned to the study, the inner validation, and the external validation cohort were 193, 130, and 96, respectively. The follow-up time of stable pGGNs for the primary and external validation cohorts were 3.66 (range, 2.01-10.08) and 4.63 (range, 2.00-9.91) years, respectively. Growth of the pGGN occurred in 166 nodules [83 (43%), 39 (30%), and 44 (45%) in the study, inner and external validation cohorts respectively]. The best-performing DLCM algorithm was DenseNet_DR, which achieved AUROCs of 0.79 [95% confidence interval (CI): 0.70, 0.86] in predicting pGGN growth in the inner validation cohort and 0.70 (95% CI: 0.60, 0.79) in the external validation cohort. Conclusions: DLCM algorithms that use chest CT images can help predict the growth of pGGNs.

Correlation between clinicopathological characteristics of lung adenocarcinoma and the risk of venous thromboembolism.

BACKGROUND: Patients with primary lung adenocarcinoma are at increased risk of venous thromboembolism (VTE). However, lung adenocarcinoma characteristics differ across histological subtypes. Therefore, we performed comprehensive analyses on the clinicopathological characteristics of lung adenocarcinoma and risk of VTE. METHODS: A total of 952 surgically resected lung adenocarcinoma cases were reviewed and classified according to criteria of the International Association for the Study of Lung Cancer (IASLC)/American Thoracic Society (ATS) /European Respiratory Society (ERS). The correlation between this classification and VTE risk was retrospectively analyzed. The risks of other clinicopathological features including pleural invasion, vascular invasion and associated surgical intervention risks were also assessed. RESULTS: Of the 952 patients, 100 (10.4%) cases experienced VTE events during the follow-up period. Among those with VTE, 28 (28%) were found before surgery, 47 (47%) were found within 1 month after surgery, and 91 (91%) were found in hospital. Univariate analysis revealed that ages, extent of resection and presence of micropapillary features were predictive of VTE risk. Furthermore, multivariable analysis demonstrated that the presence of micropapillary features (subdistribution hazard ratio [SHR] 1.560, 95% CI: 1.043-2.330) and age >60 (SHR: 2.270, 95% CI:1.491-3.470) were associated with increased risk of VTE. After one year, the probability of developing VTE was 13.1% and 8.3% in patients with micropapillary features and those without, respectively. CONCLUSIONS: VTE is a common complication for lung adenocarcinoma patients who undergo surgery, especially during the perioperative process and hospitalization. Presence of micropapillary subtype and age are positively associated with VTE risk.

Prediction modeling using routine clinical parameters to stratify survival in malignant pleural mesothelioma patients complicated with malignant pleural effusion.

BACKGROUND: Malignant pleural effusion (MPE) is common in malignant pleural mesothelioma (MPM). The survival of patients with MPM and MPE is heterogeneous. The LENT and BRIMS scores using routine clinical parameters were developed to predict the survival of patients with unselected MPE and MPM, respectively. This study aimed to stratify the survival of selected MPM patients with MPE. METHODS: Data were collected from subjects diagnosed with MPM and MPE. The LENT and BRIMS scores were applied using a combination of clinical variables to stratify subjects and compare survival characteristics. RESULTS: In total, 101 patients with MPM complicated by MPE were included in the study. The median follow-up time was 71 months (interquartile range: 24-121 months). Overall median survival was 24 (interquartile range: 12-52 months). Based on the LENT score, the low-, moderate-, and high-risk groups accounted for 65.3% (66 cases), 34.7% (35 cases), and 0%, respectively. The cumulative survival rates of the two groups were statistically significant (p = 0.031). The area under the curve (AUC) of the LENT score was 0.662. Based on the BRIMS score, the first, second, third, and fourth risk groups accounted for 1.0% (1 case), 42.9% (35 cases), 28.7% (29 cases), and 19.4% (36 cases), respectively. Survival was significantly higher in patients in the risk groups 1 and 2 than in patients in the risk groups 3 and 4 (p  = 0.037). The AUC of the BRIMS score was 0.605. CONCLUSIONS: Using routinely available clinical variables, both LENT and BRIMS scores could stratify selected MPM and MPE patients into risk groups with statistically different survival.

Predicting Axillary Lymph Node Metastasis in Early Breast Cancer Using Deep Learning on Primary Tumor Biopsy Slides.

OBJECTIVES: To develop and validate a deep learning (DL)-based primary tumor biopsy signature for predicting axillary lymph node (ALN) metastasis preoperatively in early breast cancer (EBC) patients with clinically negative ALN. METHODS: A total of 1,058 EBC patients with pathologically confirmed ALN status were enrolled from May 2010 to August 2020. A DL core-needle biopsy (DL-CNB) model was built on the attention-based multiple instance-learning (AMIL) framework to predict ALN status utilizing the DL features, which were extracted from the cancer areas of digitized whole-slide images (WSIs) of breast CNB specimens annotated by two pathologists. Accuracy, sensitivity, specificity, receiver operating characteristic (ROC) curves, and areas under the ROC curve (AUCs) were analyzed to evaluate our model. RESULTS: The best-performing DL-CNB model with VGG16_BN as the feature extractor achieved an AUC of 0.816 (95% confidence interval (CI): 0.758, 0.865) in predicting positive ALN metastasis in the independent test cohort. Furthermore, our model incorporating the clinical data, which was called DL-CNB+C, yielded the best accuracy of 0.831 (95%CI: 0.775, 0.878), especially for patients younger than 50 years (AUC: 0.918, 95%CI: 0.825, 0.971). The interpretation of DL-CNB model showed that the top signatures most predictive of ALN metastasis were characterized by the nucleus features including density (p = 0.015), circumference (p = 0.009), circularity (p = 0.010), and orientation (p = 0.012). CONCLUSION: Our study provides a novel DL-based biomarker on primary tumor CNB slides to predict the metastatic status of ALN preoperatively for patients with EBC.