I commend the authors for their review, “The Evolution of Artificial Intelligence in Medical Imaging: From Computer Science to Machine and Deep Learning,” which provides valuable insights into the advancements and applications of AI in medical imaging. The discussion of key technologies, such as convolutional neural networks (CNNs), transformers, and generative adversarial networks (GANs), highlights the field’s rapid evolution and potential for improving oncological imaging. However, several critical issues warrant attention. Claims about AI achieving or surpassing human diagnostic performance lack sufficient empirical support, as the paper does not thoroughly address limitations in the cited studies, such as small sample sizes or data biases. Additionally, the discussion of advanced methodologies, such as U-Net and transformers, is too superficial to enable replication or fully evaluate their adaptability to medical imaging challenges.
The review also falls short in critically addressing the limitations and ethical challenges of AI adoption in healthcare. Key concerns such as algorithmic bias, data imbalance, and the “black-box” nature of deep learning models are acknowledged but not explored in sufficient depth. Practical solutions, such as the use of explainability frameworks or detailed bias mitigation strategies, are missing. Furthermore, the lack of technical clarity in describing AI implementation, particularly for GANs and multimodal AI, hinders the accessibility and practical applicability of the paper. Addressing these methodological and ethical gaps would significantly bolster the review’s rigor and utility in advancing the field of medical imaging.