Diagnostic accuracy of T stage of gastric cancer from the view point of application of laparoscopic proximal gastrectomy

Deep learning model for diagnosing early gastric cancer using preoperative computed tomography images

BACKGROUND

Early gastric cancer (EGC) is defined as a lesion restricted to the mucosa or submucosa, independent of size or evidence of regional lymph node metastases. Although computed tomography (CT) is the main technique for determining the stage of gastric cancer (GC), the accuracy of CT for determining tumor invasion of EGC was still unsatisfactory by radiologists. In this research, we attempted to construct an AI model to discriminate EGC in portal venous phase CT images.

METHODS

We retrospectively collected 658 GC patients from the first affiliated hospital of Nanchang university, and divided them into training and internal validation cohorts with a ratio of 8:2. As the external validation cohort, 93 GC patients were recruited from the second affiliated hospital of Soochow university. We developed several prediction models based on various convolutional neural networks, and compared their predictive performance.

RESULTS

The deep learning model based on the ResNet101 neural network represented sufficient discrimination of EGC. In two validation cohorts, the areas under the curves (AUCs) for the receiver operating characteristic (ROC) curves were 0.993 (95% CI: 0.984-1.000) and 0.968 (95% CI: 0.935-1.000), respectively, and the accuracy was 0.946 and 0.914. Additionally, the deep learning model can also differentiate between mucosa and submucosa tumors of EGC.

CONCLUSIONS

These results suggested that deep learning classifiers have the potential to be used as a screening tool for EGC, which is crucial in the individualized treatment of EGC patients.

Clinicopathological characteristics and prognosis of poorly cohesive cell subtype of gastric cancer

BACKGROUND

The new World Health Organization (WHO) classification of gastric cancer includes a histological subtype of poorly cohesive carcinoma (PCC), which includes signet-ring cell (SRC) phenotype. We aimed to examine the concordance between preoperative clinical and postoperative histological diagnoses according to the 2010 WHO histological subtypes and to compare the prognoses of these subtypes.

METHODS

The study cohort comprised 665 patients who underwent gastrectomy from 2005 to 2019. Histological subtypes were classified into PCC-NOS (non-signet ring cell subtype), SRC, and non-PCC, which were defined by the predominant component in accordance with the 2010 WHO classification of gastric cancer. The concordance of clinical and pathological diagnosis was examined and clinicopathological characteristics and survival outcome of the three subtypes compared.

RESULTS

The cancers of 443 patients (66.7%) were classified as non-PCC, of 112 patients (16.8%) as PCC-NOS, and of 110 patients (16.5%) as SRC predominant subtypes. Significant differences in sex, age, tumor location, size, macroscopic type, and pathological TNM category (all P<0.05) were found. The concordance rate of preoperative and postoperative histological subtypes was significantly lower for poorly cohesive than other subtypes (P<0.0001). Preoperative stage tended to be underestimated for PCC-NOS subtype and these patients had poorer overall survival than those with the other two subtypes (P=0.005). Multivariate logistic regression analysis of overall survival showed that WHO histological subtype (PCC-NOS vs. non-PCC/SRC, HR: 1.64, 95% CI: 1.18-2.29, P=0.0034) was a significant independent prognostic factor.

CONCLUSION

Our results suggest that poorly cohesive carcinoma subtypes have different biological characteristics and prognoses.

Diagnostic and prognostic value of the peripheral natural killer cell levels in gastric cancer

Peripheral blood lymphocyte subsets have been reported to be useful as prognostic and/or diagnostic markers for patients with cancer. However, the clinical value of peripheral blood lymphocyte subsets in gastric cancer (GC) has remained elusive. In the present study, peripheral CD3⁺, CD4⁺ and CD8⁺ T lymphocytes, B cells (CD19⁺), regulatory T cells (Tregs; CD4⁺CD25⁺CD127^-) and natural killer (NK) cells (CD3^-CDl6⁺CD56⁺) were detected by flow cytometry in 122 patients with GC, 80 healthy donors (HDs) and 80 patients with gastric ulcer (GU). NK cells (CD56⁺) were detected by immunohistochemical (IHC) analysis in 20 GC and three GU tissue samples. A receiver-operating characteristic (ROC) curve was used to determine the threshold of the peripheral NK cell level and survival analysis was performed to assess its prognostic value in patients with GC. The results indicated that the peripheral NK cell proportion in patients with GC (18.77%) was significantly higher than that in the HD (12.19%) and GU (12.74%) groups. IHC analysis suggested that the NK level in GC tumor samples was correlated with that in paired serum samples. ROC curve analysis indicated that the peripheral NK cell level (15.16%) was able to effectively identify patients with GC, a diagnostic sensitivity of 75.41% and a specificity of 77.45% were determined. Multivariate logistic regression analysis revealed that the peripheral NK cell level was independently associated with the T stage and survival analysis demonstrated that high levels of peripheral NK cells were associated with poor prognosis of patients with GC. In conclusion, the peripheral NK cell level may be a diagnostic and prognostic marker for patients with GC.