This research proposes a method based on DenseNet for the diagnosis of diseases from medical images, with good performance in terms of accuracy and time. Therefore, the DenseNet model used to train the diverse data correctly estimated the accuracy to be around 92%. 5%, precision of 91. 7%, recall of 91. 1%, F1-score of 91. 2% and the Area Under the curve receiver operating characteristic (AUC-ROC) of 0. 942 on the test set The results In the test set, the precision at rank 942 was adequate The precision on the test set According to the results obtained above, work on improving the system's accuracy, particularly on the test set is ongoing. The comparison of the new methods to conventional diagnostic techniques also revealed certain advantages, such as higher values of performance indicators. The two aforesaid modules had an average latency of 12 ms for images of sizes 256 x 256, which makes this system a practical solution for clinical use. This research offers a possible solution to the effectiveness of diagnosis and rapidity of interpretation through an AI-motivated technique in the healthcare profession. © 2025 Elsevier B.V., All rights reserved.