Major Depressive Disorder (MDD) is a mental health condition that affects around 280 million people worldwide. Due to various lifestyle changes, many people succumb to this mental illness. Along with the growth in the medical field in diagnosing depression, there is increasing concern over the ratio of medical officer to patient, which makes the waiting period longer to detect early depression. As depression can be diagnosed through conscious or unconscious responses from patients, today, there has been considerable interest in implementing effective automated depression detection using techniques such as machine learning and deep learning (DL). Researchers have been looking for approaches to effectively identify depression in the fastest and most effective manner without jeopardizing the accuracy of the outcome. However, most research detecting MDD employs different methods, datasets, preprocessing techniques and model architectures which contributes to varying levels of accuracy. This paper aims to provide a review of the methods used for detecting depression, related public datasets and the deep learning methodology used in detecting depression from electroencephalography (EEG) signals. A taxonomy of currently available methods  for detecting depression is also provided on existing research, which helps identify a suitable method for automating depression detection. The study also explores the best possible public dataset to use for training the DL model as most of the reviewed research does not mention the meaningful parameters. Lastly, state-of-the-art deep learning models on depression detection are derived from existing studies in this area. The review is concluded with a discussion to enhance the research on depression detection using DL.