Comprehensive Summary
This study by Fan et. al, looks at brain development deviations in patients with early onset schizophrenia (EOS). This was done by applying machine learning algorithms to structural and functional MRI data. Pathological schizophrenia processes originate early in brain development and leads to brain alterations that can be detectable on MRIs and fMRIs. The data for this study was collected from 80 first episode EOS patients and 91 typically developing controls. The MRI data was collected from three imaging modalities. These modalities were then used to construct the connectomes that were being compared between the control and the EOS groups. 8 brain estimation models were first trained with the typically developing group. These models were then used to estimate brain ages for the EOS group, with the individual brain gaps being calculated by taking the difference between brain ages and chronological ages. The study found that the morphometric similarity connectome (MSC) and the structural connectome (SC) performed well when estimating brain age. The results also found that the EOS group when compared with the control group, had higher brain age gaps when using the MSC and SC features. This was correlated with the severity of their clinical symptoms.
Outcomes and Implications
This study is a good example of how machine learning algorithms can be used in neurological research studies, especially when determining a basis for neurodevelopment is necessary. The researchers mention how the brain estimation models constructed on the basis of SC and MSC features performed better than the FC-based model. This means that the connectome constructed from structural imaging is a more effective predictor of brain age than functional imaging. This knowledge can help guide future research in neurodevelopment. The researchers did state that a limitation of the study was the small population size. The small population size led to limited ability of the brain age estimation models. It must be noted that future studies must use a large enough sample size for the model to be as accurate and efficient as possible.