A machine learning approach identifies three behavioral phenotypes of TLE

Machine Learning


A group of researchers characterized three different patient groups based on behavioral patterns of temporal lobe epilepsy (TLE), highlighting the heterogeneity of this disorder.

The study, called the Epilepsy Connectome Project, included 114 patients and 83 controls. The survey results recently brain communication.

All participants will undergo experiential assessments that provide insight into behaviors such as depression, anxiety, physical problems, avoidant personality problems, attention deficit/hyperactivity, and antisocial personality problems. You have completed the inventory of the Achenbach system.

Based on analysis by unsupervised machine learning, patients who scored significantly higher overall than controls were classified into three categories of psychological risk.

  • Cluster 1: unaffected with no scale elevation compared to control
  • Cluster 2: Mild symptoms with significant elevation on several scales compared to controls
  • Cluster 3: Severely symptomatic with significant elevation on all scales compared to controls and other groups

“The person-centered analytical approach undertaken here highlights the heterogeneity of behavioral risk in patients with TLE, resulting in a more clinically meaningful group separation than traditional comparisons between TLE and controls.” commented the researchers, and to date only two other groups have taken this same approach.

The researchers compared abnormalities in measures included in the quality of life (QOL) index and the NIH toolbox emotional battery of 10 subtests of negative impact, psychological well-being, stress and self-efficacy, and social relationships. We simultaneously validated these cluster findings by identifying

Data from the NIH Toolbox Emotion Battery showed that compared to controls, patients classified in cluster 2 had significantly impaired life satisfaction (P. < .001) and self-efficacy (P. = .001), reporting a poor loneliness score (P. = .008), perceived stress (P. < .001), bodies of fear (P. < .001), angry hostility (P. = .017). Compared to cluster 1, cluster 2 patients had significantly lower scores on all but the fear effect and anger-physical aggression scales.

Patients in cluster 3 scored significantly worse compared to controls on all scales. P. = .01 Anger – Physical Aggression Scale P. < .001 on the rest of the scale.Similarly, patients in cluster 3 scored significantly worse than patients in cluster 1, with all scales P. ≤0.001.

QOL as measured by patient-weighted QOL in the Epilepsy Questionnaire Scale showed that patients classified in cluster 1 had significantly higher QOL than patients in cluster 2 on all but one scale, and patients classified in cluster 3 showed significantly worse quality of life. than patients in cluster 2 on all scales.

“These independent measures, examined as a function of the identified behavioral phenotype, reflected the same graded pattern of abnormalities reflected in the latent group,” the researchers explained. “Clearly, behavioral phenotype influences other dimensions of behavior and function. Our research shows that the identified behavioral clusters have external validity that is demonstrated through their relationship with formal psychiatric assessments.”

The researchers did not observe a clear effect of behavioral cluster type across different cognitive measures, including reading recognition tests, picture vocabulary tests, or list-sorting memory tests. patients were significantly faster than cluster 3 patients.

reference

Struck A, Garcia-Ramos C, Nair V, Presence, nature and network properties of behavioral phenotypes in other temporal lobe epilepsy. brain communicationPublished online March 30, 2023. doi:10.1093/braincomms/fcad095



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *