Parkinson’s disease (PD) traits affect millions of people the world over, with the highest prevalence in persons 50 or older. However, technical improvements were not accompanied by successful early detection of PD, making it necessary to assist clinicians in making an accurate initial diagnosis using ML-based automated methods. The objective of this study is to review in a detailed manner and compare systematically the most recent computational intelligence approaches for detection of Parkinson’s disease (PD). However, gold standard detection of PD has become largely dependent upon classification, enabling possible time savings and treatment efficiency improvements. Review of previous studies points out that there undoubtedly exists an assortment of the classification algorithms utilized in order to increase diagnose accuracy. Yet, we are still unable to identify the best classifier for PD. Therefore, as ML approaches have been employed to categorize PD patients as separate from healthy individuals or those with equivalent symptoms (i.e., other movement disorders or Parkinsonian syndromes), tools of the trade for searching the pathology–phenotype space through the lens of the phenotype (brain image) have been developed. We highlight the enormous potential of these techniques to enable a more systematic and informed PD diagnosis and assessment.