RESUMO
Advancements in artificial intelligence (AI) strengthens life-altering technology that can not only reduce human workload but also enhance speed and efficiency in medicine. Medical image segmentation, for example, MRI analysis, is an arduous task for humans that AI can accomplish efficiently. AI in medical sciences have been studied; however, current literature is deficient of state-of-the-art development in computer vision, hence, advances are needed to be addressed. In this article, we provided a novel U-net architecture which utilizes transformers, convolution neural networks (CNNs) and medical information to depict the impact of AI in medicine and cancer. This model was tested on the PROMISE-12 dataset, a complete dataset of prostate cancer MRI images, where the architecture was compared with other conventional deep learning models to evaluate its performance. Dice similarity coefficient (DSC) and loss values were used as a metric of the predicted image segmentation by the three architectures. Our data support the hypothesis that Transformer-based U-Net architecture is superior when compared to conventional AI architectures for prostate cancer MRI scans. This novel Transformer-based U-Net architecture contributes to the advancement of prostate cancer segmentation and has a future in cancer diagnosis.
