A novel decision tree approach to predict the probability of conversion to multiple sclerosis in Iranian patients with optic neuritis
Kaveh Abri Aghdam1 *, Ali Aghajani2 , Fatemeh Kanani2 , Mostafa Soltan Sanjari2 , Mehdi Moghaddasi3
- Department of Ophthalmology, Eye Research Center, The Five Senses Institute, Rassoul Akram
Hospital, Iran University of Medical Sciences, Tehran, Iran
- Department of Ophthalmology, Eye Research Center, The Five Senses Institute, Rassoul Akram hospital, Iran University of Medical Sciences, Tehran, Iran
- Department of Neurology, Hazrat Rasool Hospital, Iran University of Medical Sciences, Tehran, Iran
Abstract: Assessing the risk of conversion to multiple sclerosis (MS) in patients with optic
neuritis (ON) has been the topic of numerous studies. However, since the risk factors differ from
population to population, the extension of conclusions is a matter of debate. This study focused on
the Iranian patients with optic neuritis and assessed the probability of conversion to multiple
sclerosis by using a machine-based learning decision tree.
Methods: In this retrospective, observational study the medical records of patients with optic
neuritis from 2008 to 2018 were reviewed. Baseline vision, the treatment modality, magnetic
resonance imaging (MRI) findings, and patients’ demographics were gathered to evaluate the odds
of each factor for conversion to MS. The decision tree was then obtained from these data based on
their specificity and sensitivity to predict the probability of conversion to MS.
Results: The overall conversion rate to MS was 42.2% (117/277). 63.1 percent of patients had
abnormal MRIs at baseline. The presence of white matter plaque had the highest odds for the
conversion followed by the positive history of optic neuritis attack and gender. The regression tree
showed that the presence of plaque was the most important predicting factor that increased the
probability of conversion from 16 to 51 percent.
Conclusion: The decision tree could predict the probability of conversion to MS by considering
multiple risk factors with acceptable precision.
