Endobronchial ultrasound (EBUS) multimodal image including grey scale, blood flow doppler and elastography, can be used as non-invasive diagnosis and supplement the pathological result, which has important clinical application value. In this study, EBUS multimodal image database of 1000 inthoracic benign and malignant lymph nodes (LNs) will be constructed to train deep learning neural networks, which can automatically select representative images and diagnose LNs. Investigators will establish an artificial intelligence prediction model based on deep learning of intrathoracic LNs, and verify the model in other 300 LNs.
Intrathoracic LNs enlargement has a wide range of diseases, among which intrathoracic LNs metastasis of lung cancer is the most common malignant disease. Benign lesions, including inflammation, tuberculosis and sarcoidosis, also need to be differentiated for targeted treatment. EBUS multimodal image including grey scale, blood flow doppler and elastography, can be used as non-invasive diagnosis and supplement the pathological result, which has important clinical application value. This study includes two parts: retrospectively construction of EBUS artificial intelligence prediction model and multi-center prospectively validation of the prediction model. A total of 1300 LNs will be enrolled in the study. During the retention of videos, target LNs and peripheral vessels are examined using ultrasound hosts (EU-ME2, Olympus or Hi-vision Avius, Hitachi) equipped with elastography and doppler functions and ultrasound bronchoscopy (BF-UC260FW, Olympus or EB1970UK, Pentax). Multimodal image data of target LNs are collected. Investigators will construct artificial intelligence prediction model based on deep learning using images from 1000 LNs firstly, and verify the model in other 300 LNs. This model will be compared with traditional qualitative and quantitative evaluation methods to verify the diagnostic efficacy.
Study Type
OBSERVATIONAL
Enrollment
1,300
Shanghai Chest Hospital
Shanghai, Shanghai Municipality, China
RECRUITINGDiagnostic efficacy of EBUS multimodal artificial intelligence prediction model based on videos
Diagnostic efficacy includes sensitivity, specificity, positive predictive value, negative predictive value and diagnostic accuracy.
Time frame: 6 months post-procedure
Diagnostic efficacy of traditional qualitative and quantitative methods
Diagnostic efficacy includes sensitivity, specificity, positive predictive value, negative predictive value and diagnostic accuracy.
Time frame: 6 months post-procedure
Diagnostic efficacy of multimodal deep learning model based on images
Diagnostic efficacy includes sensitivity, specificity, positive predictive value, negative predictive value and diagnostic accuracy.
Time frame: 6 months post-procedure
Comparion of prediction model based on deeping learning with traditional qualitative and quantitative methods
Diagnostic efficacy includes sensitivity, specificity, positive predictive value, negative predictive value and diagnostic accuracy.
Time frame: 6 months post-procedure
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.