Visual gamma oscillations have been associated with low-level visual features, including contrast, colour, and size, which are highly dependent on the stimulus itself. These relationships have been characterised for artificial, isolated stimuli. Representational Similarity Analysis (RSA) has emerged as a powerful tool to correlate features or measurements from different systems that would otherwise be difficult to compare. Using natural images, we observed robust increased gamma-band activity originating bilaterally from the visual cortex (V2 region). We applied RSA to investigate possible links between neural activity in the gamma-band and image features at the level of individual fixations. Participants performed a hybrid visual and memory search task whilst eye movements and neural signals were recorded non-invasively using magnetoencephalography (MEG). For each participant, we constructed representational dissimilarity matrices (RDMs) based on gamma-band neural responses after a fixation, and convolutional neural networks (CNN) feature maps of the fixated regions. Our goal was to assess whether the structure of neural activity across fixations to different items mirrors the representational structure of visual features. Preliminary results from gamma-based RDMs show limited differentiation across fixations, highlighting challenges in linking neural responses to specific visual features.