Machine learning and artificial intelligence based Diabetes Mellitus detection and self-management: A systematic review


Diabetes Mellitus (DM) is a condition induced by unregulated diabetes that may lead to multi-organ failure in patients. Thanks to advances in machine learning and artificial intelligence, which enables the early detection and diagnosis of DM through an automated process which is more advantageous than a manual diagnosis. Currently, many articles are published on automatic DM detection, diagnosis, and self-management via machine learning and artificial intelligence techniques. This review delivers an analysis of the detection, diagnosis, and self-management techniques of DM from six different facets viz., datasets of DM, pre-processing methods, feature extraction methods, machine learning-based identification, classification, and diagnosis of DM, artificial intelligence-based intelligent DM assistant and performance measures. It also discusses the conclusions of the previous study and the importance of the results of the study. Also, three current research issues in the field of DM detection and diagnosis and self-management and personalization are listed. After a thorough screening procedure, 107 main publications from the Scopus and PubMed repositories are chosen for this study. This review provides a detailed overview of DM detection and self-management techniques which may prove valuable to the community of scientists employed in the area of automatic DM detection and self-management.


“The iridology-based or iris-based DM prediction framework is developed utilizing ML algorithms (Aminah and Saputro, 2019aAminah and Saputro, 2019b). ML is utilized to automate the detection process. The built framework comprises of methods for the creation of eye images and algorithms for image processing. Iris images are shot utilizing the Iriscope Iris Analyzer Iridology system. The Gray Level Co-Occurrence Matrix (GLCM) approach is utilized for the feature extraction to acquire the texture characteristics of the image. The KNN approach is utilized to distinguish non-diabetic and diabetic groups. Categorization tests are then checked utilizing the k-fold cross-validation process and analyzed utilizing the confusion matrix. Two classes of objects are assessed: one is 16 non-diabetic subjects and 11 are diabetic subjects. The findings reveal that the precision is 85.6% the false-positive rate (FPR) is 11.07%, the false-negative rate (FNR) is 20.40%, the specificity is 0.889, and the sensitivity is 0.796.”

Authors: JyotismitaChakiaS.Thillai GaneshbS.KCidhambS.Ananda Theertanb

School of Information Technology and Engineering, Vellore Institute of Technology, Vellore, India
School of Electronics Engineering, Vellore Institute of Technology, Vellore, India

Download full abstract: 1-s2.0-S1319157820304134-main