Medical image intelligent segmentation algorithms based on deep learning (DL) can help doctors accurately locate and recognize disease areas, thereby improving diagnostic accuracy and treatment effectiveness. The intelligent segmentation algorithm utilizes Convolutional Neural Network (CNN) DL models to train preprocessed medical images. Subsequently, the model was subjected to k-fold cross validation evaluation and random gradient descent parameter optimization. Finally, the trained model was utilized to compare the accuracy of medical images with traditional methods, and segmentation experiments were conducted by calculating Dice coefficients and recall rates. The results showed that the Dice coefficient of the CNN model was around 0.95, and the recall rate approached 1 faster. Therefore, the intelligent segmentation algorithm for medical images based on DL technology can more accurately identify and locate diseases and tumors, improving the accuracy of doctor diagnosis and treatment effectiveness.