NeuroRegulation
September 29, 2022
Pankaj Pandey, Pragati Gupta, Krishna Prasad Miyapuram
10 citations
Comparing three meditation traditions—Himalayan Yoga, Isha Shoonya, and Vipassana—using EEG to build functional brain networks and graph measures reveals distinct neural patterns. Classifying traditions against a control group with support vector machines achieved up to 90% accuracy (alpha band for Isha Shoonya). Key findings include higher delta connectivity in Vipassana meditators, stronger synchronization of left anterior frontal theta networks across traditions, greater gamma2 processing in Himalayan and Vipassana meditators, increased left frontal activity in theta and gamma bands for all meditators, and extensive modularity in gamma processing. The work suggests implications for neurotechnology to guide novice meditators.
Fitoterapia
June 1, 2023
Caroline V L Moreira, Ana Luiza G Faria, Daiany P B Silva et al.
2 citations
A new compound called P-3l, an analogue of salvinorin A, reduces pain and anxiety-like behaviors in mice after oral administration at doses of 1, 3, 10, and 30 mg/kg. It lessened acetic acid-induced writhing, formalin-induced paw licking, hotplate responses, and aversion in elevated plus-maze, open field, and light-dark box tests. P-3l also boosted the effects of morphine and diazepam at low doses without causing changes in organ weight or blood parameters. The pain and anxiety relief were blocked by naloxone, naloxonazine, nor-binaltorphimine, and flumazenil, indicating involvement of opioid receptors and the benzodiazepine site. These findings suggest P-3l may have clinical potential for pain and anxiety treatment.
arXiv Preprint Archive
June 12, 2021
Pankaj Pandey, Krishna Prasad Miyapuram
A lightweight convolutional neural network (CNN) is proposed to classify cognitive states from EEG recordings. The pipeline first converts neural time-series signals into 2D spectral images that preserve electrode relationships and spectral properties, then uses a network with standard, depth-wise, and separable convolutions to reduce parameters. Tested on an open-access meditation dataset with expert, nonexpert, and control states, the model achieves comparable classification performance to six machine learning classifiers and four deep learning models while using less than 4% of their parameters, making it suitable for real-time neurofeedback.
Pankaj Pandey, Julio Rodriguez-Larios, Krishna Prasad Miyapuram et al.
Machine learning models can detect moments of distraction during meditation by analyzing EEG signals. Using data from 24 novice meditators performing breath focus meditation, researchers extracted twelve linear and non-linear EEG features and tested ten supervised classifiers. Linear features achieved up to 86% accuracy in distinguishing awake from sleepy states, while non-linear features reached nearly 78% accuracy in distinguishing awake from mind-wandering states. Unsupervised t-SNE visualization confirmed distinct clusters for each condition. These findings support the development of mobile EEG neurofeedback protocols that alert meditators when they become distracted.