Evaluation of Submucosal or Lymphovascular Invasion Detection Rates in Early Gastric Cancer Based on Pathology Section Interval

An artificial intelligence system for comprehensive pathologic outcome prediction in early gastric cancer through endoscopic image analysis (with video)

BACKGROUND

Accurate prediction of pathologic results for early gastric cancer (EGC) based on endoscopic findings is essential in deciding between endoscopic and surgical resection. This study aimed to develop an artificial intelligence (AI) model to assess comprehensive pathologic characteristics of EGC using white-light endoscopic images and videos.

METHODS

To train the model, we retrospectively collected 4,336 images and prospectively included 153 videos from patients with EGC who underwent endoscopic or surgical resection. The performance of the model was tested and compared to that of 16 endoscopists (nine experts and seven novices) using a mutually exclusive set of 260 images and 10 videos. Finally, we conducted external validation using 436 images and 89 videos from another institution.

RESULTS

After training, the model achieved predictive accuracies of 89.7% for undifferentiated histology, 88.0% for submucosal invasion, 87.9% for lymphovascular invasion (LVI), and 92.7% for lymph node metastasis (LNM), using endoscopic videos. The area under the curve values of the model were 0.992 for undifferentiated histology, 0.902 for submucosal invasion, 0.706 for LVI, and 0.680 for LNM in the test. In addition, the model showed significantly higher accuracy than the experts in predicting undifferentiated histology (92.7% vs. 71.6%), submucosal invasion (87.3% vs. 72.6%), and LNM (87.7% vs. 72.3%). The external validation showed accuracies of 75.6% and 71.9% for undifferentiated histology and submucosal invasion, respectively.

CONCLUSIONS

AI may assist endoscopists with high predictive performance for differentiation status and invasion depth of EGC. Further research is needed to improve the detection of LVI and LNM.

A Modified eCura System to Stratify the Risk of Lymph Node Metastasis in Undifferentiated-Type Early Gastric Cancer After Endoscopic Resection

PURPOSE

The original eCura system was designed to stratify the risk of lymph node metastasis (LNM) after endoscopic resection (ER) in patients with early gastric cancer (EGC). We assessed the effectiveness of a modified eCura system for reflecting the characteristics of undifferentiated-type (UD)-EGC.

MATERIALS AND METHODS

Six hundred thirty-four patients who underwent non-curative ER for UD-EGC and received either additional surgery (radical surgery group; n=270) or no further treatment (no additional treatment group; n=364) from 18 institutions between 2005 and 2015 were retrospectively included in this study. The eCuraU system assigned 1 point each for tumors >20 mm in size, ulceration, positive vertical margin, and submucosal invasion <500 µm; 2 points for submucosal invasion ≥500 µm; and 3 points for lymphovascular invasion.

RESULTS

LNM rates in the radical surgery group were 1.1%, 5.4%, and 13.3% for the low- (0-1 point), intermediate- (2-3 points), and high-risk (4-8 points), respectively (P-for-trend<0.001). The eCuraU system showed a significantly higher probability of identifying patients with LNM as high-risk than the eCura system (66.7% vs. 22.2%; McNemar P<0.001). In the no additional treatment group, overall survival (93.4%, 87.2%, and 67.6% at 5 years) and cancer-specific survival (99.6%, 98.9%, and 92.9% at 5 years) differed significantly among the low-, intermediate-, and high-risk categories, respectively (both P<0.001). In the high-risk category, surgery outperformed no treatment in terms of overall mortality (hazard ratio, 3.26; P=0.015).

CONCLUSIONS

The eCuraU system stratified the risk of LNM in patients with UD-EGC after ER. It is strongly recommended that high-risk patients undergo additional surgery.

Prognostic Implication of Lymphovascular Invasion in Early Gastric Cancer Meeting Endoscopic Submucosal Dissection Criteria: Insights from Radical Surgery Outcomes

BACKGROUND

The management of early gastric cancer (EGC) has witnessed a rise in the utilization of endoscopic submucosal dissection (ESD) as a treatment modality, although prognostic markers are needed to guide management strategies. This study investigates the prognostic implications of lymphovascular invasion (LVI) in ESD-eligible EGC patients, specifically its implications for subsequent radical surgery.

MATERIAL AND METHODS

A retrospective, multicenter study from two primary hospitals analyzed clinicopathological data from 1369 EGC patients eligible for ESD, who underwent gastrectomy at Shanghai Cancer Center and Huashan Hospital between 2009 and 2018. We evaluated the relationship between LVI and lymph node metastasis (LNM), as well as the influence of LVI on recurrence-free survival (RFS) and overall survival (OS).

RESULTS

We found a strong association between LVI and LNM (p < 0.001). Advanced machine learning approaches, including Random Forest, Gradient Boosting Machine, and eXtreme Gradient Boosting, confirmed the pivotal role of LVI in forecasting LNM from both centers. Multivariate analysis identified LVI as an independent negative prognostic factor for both RFS and OS, with hazard ratios of 4.5 (95% CI: 2.4-8.5, p < 0.001) and 4.4 (95% CI: 2.1-8.9, p < 0.001), respectively.

CONCLUSIONS

LVI is crucial for risk stratification in ESD-eligible EGC patients, underscoring the necessity for radical gastrectomy. Future research should explore the potential incorporation of LVI status into existing TNM staging systems and novel therapeutic strategies.

Ensemble Deep Learning Model to Predict Lymphovascular Invasion in Gastric Cancer

Lymphovascular invasion (LVI) is one of the most important prognostic factors in gastric cancer as it indicates a higher likelihood of lymph node metastasis and poorer overall outcome for the patient. Despite its importance, the detection of LVI(+) in histopathology specimens of gastric cancer can be a challenging task for pathologists as invasion can be subtle and difficult to discern. Herein, we propose a deep learning-based LVI(+) detection method using H&E-stained whole-slide images. The ConViT model showed the best performance in terms of both AUROC and AURPC among the classification models (AUROC: 0.9796; AUPRC: 0.9648). The AUROC and AUPRC of YOLOX computed based on the augmented patch-level confidence score were slightly lower (AUROC: -0.0094; AUPRC: -0.0225) than those of the ConViT classification model. With weighted averaging of the patch-level confidence scores, the ensemble model exhibited the best AUROC, AUPRC, and F1 scores of 0.9880, 0.9769, and 0.9280, respectively. The proposed model is expected to contribute to precision medicine by potentially saving examination-related time and labor and reducing disagreements among pathologists.

A robust model training strategy using hard negative mining in a weakly labeled dataset for lymphatic invasion in gastric cancer

Gastric cancer is a significant public health concern, emphasizing the need for accurate evaluation of lymphatic invasion (LI) for determining prognosis and treatment options. However, this task is time-consuming, labor-intensive, and prone to intra- and interobserver variability. Furthermore, the scarcity of annotated data presents a challenge, particularly in the field of digital pathology. Therefore, there is a demand for an accurate and objective method to detect LI using a small dataset, benefiting pathologists. In this study, we trained convolutional neural networks to classify LI using a four-step training process: (1) weak model training, (2) identification of false positives, (3) hard negative mining in a weakly labeled dataset, and (4) strong model training. To overcome the lack of annotated datasets, we applied a hard negative mining approach in a weakly labeled dataset, which contained only final diagnostic information, resembling the typical data found in hospital databases, and improved classification performance. Ablation studies were performed to simulate the lack of datasets and severely unbalanced datasets, further confirming the effectiveness of our proposed approach. Notably, our results demonstrated that, despite the small number of annotated datasets, efficient training was achievable, with the potential to extend to other image classification approaches used in medicine.

External Validation of the eCura System for Undifferentiated-Type Early Gastric Cancer with Noncurative Endoscopic Resection

Background/Aims

The eCura system, a scoring model for stratifying the lymph node metastasis risk after noncurative endoscopic resection for early gastric cancer (EGC), has been internally validated, primarily for differentiated-type EGC. We aimed to externally validate this model for undifferentiated-type EGC.

Methods

This multicenter, retrospective cohort study included 634 patients who underwent additional surgery (radical surgery group, n=270) or were followed up without additional treatment (no additional treatment group, n=364) after noncurative endoscopic resection for undifferentiated-type EGC between 2005 and 2015. The lymph node metastasis and survival rates were compared according to the risk categories.

Results

For the radical surgery group, the lymph node metastasis rates were 2.6%, 10.9%, and 14.8% for the low-, intermediate-, and high-risk eCura categories, respectively (p for trend=0.003). For the low-, intermediate-, and high-risk categories in the no additional treatment group, the overall survival (92.7%, 68.9%, and 80.0% at 5 years, respectively, p<0.001) and cancer-specific survival rates (99.7%, 94.7%, and 80.0% at 5 years, respectively, p<0.001) differed significantly. In the multivariate analysis, the hazard ratios (95% confidence interval) in the no additional treatment group relative to the radical surgery group were 3.18 (1.41 to 7.17; p=0.005) for overall mortality and 2.60 (0.46 to 14.66; p=0.280) for cancer-specific mortality in the intermediate-to-high risk category. No such differences were noted in the low-risk category.

Conclusions

The eCura system can be applied to undifferentiated-type EGC. Close follow-up without additional treatment might be considered for low-risk patients, while additional surgery is recommended for intermediate- and high-risk patients.

Comparison of long-term outcomes of endoscopic submucosal dissection and surgery for undifferentiated-type early gastric cancer meeting the expanded criteria: a systematic review and meta-analysis

BACKGROUND

There have been concerns over the long-term outcomes of endoscopic submucosal dissection (ESD) for undifferentiated-type early gastric cancer (UD EGC). We aimed to compare the long-term outcomes of ESD and surgery for patients with UD EGC.

METHODS

We searched PubMed, Embase, and Cochrane Library databases through March 2021 to identify studies that compared the long-term outcomes of ESD and surgery for UD EGC meeting expanded criteria for curative resection. The risk of bias was assessed with the Cochrane tool for non-randomized studies. The risk ratio (RR) was estimated using a fixed-effect model.

RESULTS

Overall, 1863 patients from five retrospective cohort studies, including 908 patients with propensity score matching (PSM), were eligible for meta-analysis. ESD was associated with inferior overall survival (OS) compared to surgery in the overall cohort (RR 2.11; 95% CI 1.26-3.55) but not in the PSM cohort (RR 1.18; 95% CI 0.60-2.32). In the PSM cohort, ESD had a lower disease-free survival (DFS) (RR 2.49; 95% CI 1.42-4.35) and higher recurrence (RR 12.61; 95% CI 3.43-46.37), gastric recurrence (RR 11.25; 95% CI 3.06-41.40), and extragastric recurrence (RR 4.23; 95% CI 0.47-37.93). Recurrence outcomes were similar between the overall and PSM cohorts. Disease-specific survival was not significantly different between the two groups in both the overall and PSM cohorts.

CONCLUSION

Although OS after curative ESD for UD EGC was not different from that after surgery in the PSM cohort, DFS and recurrence were inferior after ESD. Limitations included a lack of randomized trials. Further prospective studies comparing the long-term outcomes of ESD and surgery for UD EGC are needed (PROSPERO CRD 42021237097).

Therapeutic approach to non-curative resection after endoscopic treatment in early gastric cancer

Background: Endoscopic resection is indicated for early or superficial gastrointestinal neoplasms with a negligible risk of lymph node metastasis. This procedure could preserve the organ while allowing en bloc resection of tumors, irrespective of the size and location of the lesion. Histological evaluation of the resected specimen determines whether curative resection, which implies a favorable long-term outcome, was achieved. If the resected specimen reveals non-curative, additional treatment is necessary as it is strongly associated with recurrence.Current Concepts: Surgical resection is recommended after non-curative resection of gastrointestinal neoplasms. However, rather than surgical resection, additional endoscopic treatment can be recommended if non-curative resection is solely because of the positive involvement at the horizontal resection margin without any other findings compatible with the non-curative resection criteria.Discussion and Conclusion: Adopting precise indications of endoscopic resection is important to reduce the risk of non-curative resection. If curative resection is not achieved after endoscopic resection, additional treatment should be considered to prevent local recurrence as well as lymph node metastasis.

Comparison between endoscopic submucosal resection and surgery for the curative resection of undifferentiated-type early gastric cancer within expanded indications: a nationwide multi-center study

BACKGROUND AND AIMS

Endoscopic submucosal dissection (ESD) for undifferentiated early gastric cancer (UD EGC) has debate due to the risk of lymph node metastasis. We investigated the outcomes of ESD compared to those of surgery for the UD EGC within expanded indication.

METHODS

We reviewed 971 UD EGC patients performed ESD across 18 hospitals in Korea and 1812 patients who underwent surgical resection in two hospitals between February 2005 and May 2015. Of these cases, we enrolled a curative resected ESD group of 328 patients and surgery group of 383 cases within an expanded indication. Overall outcomes and one-to-one propensity score-matched (218 ESD group vs 218 surgery group cases) outcomes for these two groups were analyzed.

RESULTS

Over the 75.6 month median follow-up period for the 711 enrolled cases, recurrences occurred in 22 patients (6.7%) in the ESD group but not in the surgery group. Overall survival (OS) was higher in the surgery group (p = 0.0316) in all cases, but there was no significant difference after propensity score matching (p = 0.069). According to the histologic type in propensity score matching, the OS of signet ring cell carcinoma and poorly differentiated carcinoma patients did not differ between the ESD and surgery groups (p = 0.1189 and p = 0.3087, respectively). In the surgery group involving expanded criteria, lymph node metastasis was found in six cases (1.56%).

CONCLUSIONS

Although ESD shows comparable outcomes to surgery for the UD EGC within expanded indications, appropriate patient selection is needed for the ESD due to the possibility of lymph node metastasis.

Discrepancy between endoscopic and pathological ulcerative findings in clinical intramucosal early gastric cancer

BACKGROUND

Ulcerative finding (UL) is one of the factors that define the indication and curability of endoscopic resection (ER) in early gastric cancer (EGC). Discrepancies between endoscopic UL (cUL) and pathological UL (pUL) sometimes occur in clinical practice. The aim of this study was to investigate the discrepancy rate in UL diagnosis and the risk factors associated with such discrepancies.

METHODS

Patients with clinical intramucosal (cT1a) EGC who underwent ER or surgery between September 2002 and December 2017 were analyzed. The proportion of cUL-negative (cUL0) lesions that were identified as pUL-positive (pUL1) and that of cUL-positive (cUL1) lesions that were identified as pUL-negative (pUL0) were calculated. Logistic regression analysis was performed to estimate the associations between discrepancy in UL diagnosis and clinical variables of the lesion, such as the size, histology, location, and macroscopic type.

RESULTS

In total, 5382 lesions were evaluated; 5.5% of cUL0 lesions (256/4619) were identified as pUL1, while 38.7% of cUL1 lesions (295/763) were pUL0. Multivariate analysis indicated that in cUL1 lesions, tumor location in the lower third of the stomach (odds ratio 3.11, 95% confidence interval 1.90-5.08) was identified as an independent risk factor for overestimation.

CONCLUSIONS

Endoscopic diagnosis of UL in cT1a EGC was overestimated in 38.7% of lesions, especially for lesions located in the lower third of the stomach. This discrepancy should be considered in the management of cT1a EGC with UL.