This study aimed to non-invasively detect electromyography (EMG) activity of deep thumb muscles using surface electrodes. Researchers placed electrodes on the forearms of 15 participants and recorded EMG signals while participants performed thumb movements. Independent component analysis was used to separate EMG signals from superficial and deep muscles. Predicted EMG waveforms for each deep muscle were correlated with independent components, and the highest correlated component was considered to represent that muscle's activity. Overall high correlations were found between predicted and recorded waveforms. Accuracy, sensitivity and specificity measures between predicted and recorded waveforms were also statistically significant when using a threshold activation level, demonstrating the first non-invasive detection of EMG activity from deep thumb muscles.