Is it possible to identify chronic fatigue syndrome via machine learning techniques?

Background: Chronic fatigue syndrome (CFS) is characterized by persistent exhaustion lasting at least six months, post-exercise malaise, discomfort, and neurological and autonomic symptoms. Numerous efforts to formulate a standardized definition for CFS have been made and the most extensive diagnostic tools are Fukuda criteria and Canadian Consensus Document. Since CFS symptoms overlap other disorders, identifying CFS has proven difficult. Therefore, the development of accurate diagnostic tools, such as Machine Learning (ML) techniques, is necessary. Neuroimaging and psychometric testing may be augmented using ML in different fields such as malingering, and clinical medicine. ML employs algorithms and statistical models to draw inferences from patterns in data. Herein, we provide an overview of the state of the art on the current ML protocols used to diagnose CFS. Methods: We conducted a literature search of available sources on ML protocols to diagnose CFS. Results: Watson and colleagues demonstrated that DePaul Symptom Questionnaire (DSQ) data in combination of ML techniques can provide accurate basis for diagnosing CFS; furthermore, Provenzano and colleagues implemented a ML predictive model to diagnose CFS by basing on magnetic resonance imaging structural and functional data. Conclusion: ML models have shown promising advantages in solving classification problems, addressing clinical questions such as identifying predictors that distinguish patients from healthy controls, thus allowing to make inferences at group and individual level. Future research is needed to substantiate the aforementioned findings, and to develop models able to perform differential diagnosis by identifying key pathognomonic cluster of neuroimaging, immunological and behavioral characteristics of CFS.