Comparison of Prostate Cell Image Classification Using CNN; Xception and DenseNet-201

Prostate cancer is the second highest cause of cancer deaths in men in the United States, with an estimated 34,700 deaths from this disease in 2023. Until now, the causes of this cancer are still unknown compared to other cancers. Several factors that can influence cancer risk include age, ethnicity, genetic factors, and family history. In detecting this disease there are many problems such as lack of accurate diagnosis. To overcome this problem, this research will use artificial intelligence to classify 7 images of prostate cells. The artificial intelligence method used is the CNN method from deep learning which is applied to images of prostate cells. The two CNN models used are Xception and DenseNet-201. There are two tests in this research, namely at the training stage and the testing stage. As a result, at the training stage Xception had higher accuracy results (99.58%) than DenseNet-201 (98.85%). At the training stage, an analysis of the time required for the model in the classification process is also carried out. This process, DenseNet-201 took longer (with an average of 95 minutes) than Xception (with an average of 69 minutes 24 seconds). Apart from training, analysis was also carried out at the testing stage, where Xception also outperformed at this stage with slightly higher accuracy than DenseNet-201 (97.44% (STD ±1.26)>97.32% (STD ±1.49)). Overall, the analysis shows that Xception provides better performance in training and testing than DenseNet-201.