Reaching hand movement is an extensively studied motor skill in brain-computer interface (BCI). Past studies have shown that neural activity during movement can be used to predict executed hand movements with great success. To truly benefit the end-users of BCI, however, development of decoding strategies to predict movement in the absence of actual movement is highly necessary, as the people who are likely to benefit the most from BCI systems are patients who suffer from immobility. As such, we came to question whether brain signals during motor imagery could be decoded to predict continuous hand trajectories in a reliable manner. Here, we designed a study to investigate the feasibility of decoding imagined trajectories of hand movement by selecting movement-related neural features distributed in distinct spectral, spatial and temporal coordinates during movement imagination. We selected electrocorticography (ECoG) as our choice of neural signal due to its superior spatiotemporal resolution over the widely used electroencephalography (EEG). Two epileptic patients volunteered to participate in this study and performed both imagination and execution of reaching hand movements to four distinct target locations in three-dimensional space. Raw ECoG signals were processed to three different types of features – amplitude modulation (AM), power spectral density (PSD), and a feature we named low-pass filtered ECoG potential (LFEP). Selected features from different feature types were used to train a Bayesian regression-based decoding model to predict imagined trajectories of hand movement. We found that continuous hand trajectory during imagined movement could be decoded successfully, with decoding accuracies achieved at values higher than those obtained from EEG. Analysis of the spectral distribution of neural features revealed that frequency bands in the high gamma range (80>Hz) were most useful in decoding, which coincided with the finding from spatial distributions showing active involvement of the motor cortex in both imagined and executed reaching movements. The results suggest that decoding imagined hand trajectories is highly possible using ECoG signals in humans and serves as evidence to support choice of ECoG as a suitable candidate for developing movement-free BCI.