A deep detection network to quantify the vascular morphological distribution of esophageal mucosal IPCLs

Abstract

The invention belongs to the technical field of medical image processing, in particular to a depth detection network for quantifying the morphological distribution of esophageal mucosal IPCLs blood vessels. The present invention includes feature extraction network and feature pyramid, region candidate network, interest region pooling and cluster distribution prior self-embedded cancer focus classification network, and visualization system on narrow-band imaging endoscopic images. The feature extraction network extracts the feature map of the input image; the feature pyramid fuses the features of different scales; the region candidate network proposes possible lesion regions; the pooling of the interest region pools the features to the suspicious lesion region; the cluster distribution prior self-embedding The cancer foci classification network of the system classifies the cancer foci; finally, it is visualized on the narrow-band imaging endoscopic image, and the cancer foci are box-marked with different colors. The present invention detects and diagnoses the cancer foci of early esophageal squamous cell carcinoma existing in the image, which can effectively improve the diagnosis efficiency and assist the doctor to obtain higher diagnosis accuracy.

Description

technical field [0001] The invention belongs to the technical field of medical image processing, and in particular relates to a depth detection network for quantifying the morphological distribution of esophageal mucosal IPCLs blood vessels. Background technique [0002] Esophageal cancer and gastric cancer have poor prognosis, with 5-year relative survival rates of 20.9% and 27.4%, respectively, bringing a serious burden to health care [11,13-14] . Standardized upper gastrointestinal cancer screening, treatment, and follow-up are effective means to reduce cancer morbidity and mortality. Among them, narrow-band imaging endoscopic screening is the first choice for the detection of upper gastrointestinal cancer. The pathological type and infiltration depth of esophageal mucosal lesions under narrow-band imaging endoscopy are mainly based on their unique vascular morphology of intrapapillary capillary loops (IPCLs). [0003] According to the classification criteria proposed ...

AI Technical Summary

Problems solved by technology

In this case, clinicians need to observe all the structures, which is extremely easy to cause visual fatigue. Coupled with the lack of clinical experience, after observing 5-10 visual fields, clinicians often only remember the impression "Refer to Murphy's Law" for particularly profound parts, lacking an objective and quantifiable concept, which may easily lead to misjudgment of the condition and errors in medical decision-making

Images