(Learn more about this study)
Initial CDAS Request Approval
Apr 16, 2020
Development of a Novel Prognostic Model for Predicting Metastasis using CNN method in Colorectal Cancer
Colorectal cancer (CRC) is the third most common cancer and the fourth main cause of cancer death worldwide, its importance is expected to increase significantly over the next years. In recent years, the number of dysplastic and early CRC lesions has increased with increasing awareness and the introduction of screening and surveillance programs for CRC. Early CRC is a tumor that is confined to the mucosa or submucosa and does not invade the muscularis propria. In certain cases of early CRC, endoscopic resection is the standard treatment that is equal to surgery, with the advantages of less invasive and less expensive. However, in the presence of lymph node metastasis (LNM) or distant metastasis, cancer cannot be treated with local endoscopic resection alone; it requires lymphadenectomy to minimize the risk of local recurrence. Patients with submucosal invasive CRC have LNM in 6–16% of cases. The number of patients with LNM is presumably underestimated, as clinicians make important treatment decisions based on limited examinations, such as ultrasound and computed tomography (CT). Thus, a fast and accurate assessment of the risk of locoregional LNM after local endoscopic treatment of patients with early CRC is essential to determine whether these patients should undergo additional surgery or be monitored regularly. Currently, no universally accepted indications and criteria exist for additional surgical resection after endoscopic resection, even though a fast and accurate assessment of the risk of locoregional LNM after local endoscopic treatment of patients with early CRC necessary. Tumor tissue image scanning is becoming part of routine clinical practice for the acquisition of high resolution tumor histological details. In recent years, several computer algorithms for hematoxylin and eosin (H&E) stained pathology image analysis have been developed to aid pathologists in objective clinical diagnosis and prognosis. Examples include an algorithm to extract stromal features and an algorithm to assess cellular heterogeneity as a prognostic factor in colon cancer. More recently, studies have shown that morphological features are associated with patient prognosis in colon cancer as well. Deep learning methods, such as convolutional neural networks (CNNs), have been widely used in image segmentation, object classification and recognition.
Therefore, the purpose of this study was to identify a novel predictive model for metastasis by using simple clinical and histopathological parameters with high reliability, that could be used to better stratify patients with early CRC for lymph node or distant metastasis.
We aimed to develop an accurate predictive model for lymph node or distant metastasis using histopathological image and clinical parameters in early CRC.