Migraine place a significant burden on health and economies, and it is typically diagnosed using symptom-based clinical histories, which suffers delays or incorrectly diagnosed in difficult cases. Recent advances in artificial intelligence (AI), machine learning (ML) and deep learning (DL), which allow automated analysis of neuroimaging and multimodal clinical data to find early markers and sub-types. This paper describes a review of 25 contributions to the literature applying DL to neuroimaging and ML to multimodal integration, of the last decade, comparing modalities, architectures, and performance. CNNs on rs-fMRI and 3D deep achieve accuracies as high as 99.25% but multimodal models perform better across populations. The review pointed out the latest advanced techniques, main obstacles (small data, absence of standardization), and indicated future research trends concerning explainable AI frameworks to be used in clinical practice.