Applications of Intelligent Computer Diagnosis Systems in Different Majors of Medicine

Background and aims: As artificial intelligence (AI) technics have dramatically grown, a newhorizon is appeared in the field of medicine called intelligent medical diagnosis. Diagnosis systemshave developed in all majors of medicine like neurology, psychiatry, anesthesiology, epidemiology,pathology, gynecology and obstetrics, radiology, cardiology and genetic. Early detectionof some critical diseases, decision making in situations of confronting high number of medicalfactors, differentiating of patients who have similar signs and symptoms but they have differentdiseases, detecting of variations in medical images that cannot be seen by visual inspections, predictingthe prevalence of contagious disease, are the examples of intelligent diagnosis systems.Method: There are several known AI methods for the diagnosis in medical data. Some of thesemethods along with their applications are explained as follows: Mathematical inverse problemfor estimating the localization of a lesion (e.g., focal source of seizure); Using graph theory tofind functional connectivity among different parts of the brain (e.g., differentiating of Parkinson,Alzheimer, MCI diseases, pain level detection). Applying deep neural networks for classificationand regression in big data analysis (e.g., depression level detection, seizure detection, seizure prediction,sleep stage classification); Using fuzzy logic inference systems in confrontation of situationsincluding uncertainty in data (e.g., missing features); Mapping EEG signals to the Riemannianspace for better separation of patients (e.g., BMD and Schizophrenia); Applying statisticalmachine learning methods and pattern recognition methods to classify the signals and images ofdifferent patients with similar clinical signs. To customize a model and optimize its parameters fora specific medical data, evolutionary-based search methods are used. When the labels of samplesare continuous, statistical and neural regression models can be used to determine continuous scoreof the diseases, which have been previously determined by filling up a questionnaire.Results: In this part, a review over the previous results is presented. High rate classificationaccuracy of patients with attention deficit hyper activity disorder from bipolar manic depression(BMD) patients by characterizing their electroencephalogram (EEG) signals both in the idle stateand in presence of visual stimuli were achieved. In addition, BMD cases were differentiated fromschizophrenic patients by analyzing their steady state visual evoke potentials. Localization offocal seizure sources was estimated with high precision in the range of millimeter by differentresearch teams. Electromyogram signals were repeatedly diagnosed to estimate motor unit actionpotential shapes and firing rate of motor units. Covid-۱۹ death rate was frequently estimated bystatistical regression methods. Moreover, computed tomography images of patients with Covid-۱۹ were applied to deep neural networks and the diagnosis precision was reported convincingby several reams. In continue of our reviewing, infertility rate was successfully predicted by datamining methods considering efficient factors. Plenty of electrocardiography signals belonged todifferent heart disease (e.g., ventricular tachycardia and ventricular fibrillation) were diagnosedand detected. Super-family proteins are quiet high dimensional data and classified by fuzzy networks.Conclusion: Some of intelligent diagnosis systems are commercialized in the fields of cardiology,radiology and neurology. Early diagnosis of some diseases like Alzheimer by these intelligentsystems can significantly prolong the patients’ life. Nevertheless, some of these diagnosis systemsare still spending their infancy period and still could not provide promising results.

کلیدواژه ها:

artificial intelligence ، medical diagnosis systems ، early detection ، neurology and psychiatry ، rehabilitation