For the past few decades, intracranial hemorrhage (ICH) accounts for half of the disability-adjusted life years lost because of the stroke. The therapy of acute stroke remains difficult since no one medication has been proven to enhance the prognosis. The subject of artificial intelligence (AI) is fast emerging, through new prospects in diagnostic radiology being made probable via development of electronic health data and contemporary technology advancements. Deep learning (DL) algorithms have demonstrated consistent improvement in performance on a range of medical image tasks in modern trends. DL methods have been suggested as a technique for non-contrast computed tomography (NCCT) head imaging to identify different types of cranial bleeding. The ability of the DL algorithm to enable image interpretation in subtle, acute cases could increase the diagnostic production of computed tomography (CT) for the identification of this urgent condition, perhaps accelerating treatment when essential and enhancing the outcomes of patients. Even though, there is a tendency to postpone the early diagnosis of ICHs due to the extensive use of CT scans. Currently, numerous models are exploited to diagnose the brain hemorrhage and tumors. Still, the deep extraction and appropriate training models have crucial effects on the extraction of CT image features to detect brain hemorrhages. Furthermore, the adoption of deep learning algorithms faces several obstacles, including the lack of labeled datasets, generating algorithms that analyze medical images volumetrically, and the intricate logistics of integrating the algorithms into clinical settings. Overall, the literature and methods used in the development of DL algorithms for the identification of cranial hemorrhage on NCCT head investigations are descriptively examined in this analysis.