IDDF2024-ABS-0068 Constructing a predictive model for tumor staging of colorectal cancer under endoscopic ultrasound based on deep convolutional neural networks

<h3>Background</h3> To develop an artificial intelligence diagnostic model based on a convolutional neural network (CNN). It can automatically identify the lesion area of colorectal cancer under endoscopic ultrasound and predict the tumor staging of colorectal cancer under endoscopic ultrasound. This method avoids staging errors caused by human factors and increases the objectivity and uniformity of endoscopic ultrasound in identifying colorectal cancer lesions and staging of colorectal cancer. <h3>Methods</h3> This study included 1436 images of patients with colorectal cancer who underwent endoscopic ultrasound at the Endoscopy Center of Foshan Second People’s Hospital from February 2022 to February 2024. Randomly divide these images into a training set and a testing set. Using the CNN algorithm as the basic architecture, use the training set to construct a prediction model for the lesion and non-lesion areas of colorectal cancer under endoscopic ultrasound, as well as its staging. Use the testing set images to validate the model and evaluate the diagnostic effectiveness of the prediction model. <h3>Results</h3> A total of 1436 images were included in this study, including 1275 in the training set and 161 in the testing set. In the training set, there are 84, 273, 806, and 112 form T1 to T4 stages, respectively. In the test set, this model can effectively distinguish between lesion and non-lesion areas, with an accuracy of 82.1%. The accuracy of T2 and T3 in this prediction model is 82.4% and 89.3%, respectively. <h3>Conclusions</h3> This study constructed a CNN model for identifying the lesion and non-lesion areas of colorectal cancer under endoscopic ultrasound, as well as a predictive model for T staging. This model not only effectively distinguishes the lesion and non-lesion areas, but also accurately predicts their T staging. This model can help clinical endoscopic ultrasound physicians quickly identify colorectal cancer and provide a reference value for T staging, but its stability and practicality still need further testing.