Neuroprosthetics equipped with artificial synapses hold promise to address some most intricate medical problems, such as human sensory disorders. Yet, it is necessitated and of paramount importance for neuroprosthetics to be able to differentiate significant and insignificant signals. Here, we present an information-filterable artificial retina system that integrates artificial synapses with a signal-integration device for signal perception and processing with attention. The synaptic weight modulation is rendered through metal-organic framework (MOF) layers, where distinct short-term and long-term properties are predominantly determined by MOF's pore diameter and functionality. Specifically, four types of isoreticular Zr-based MOFs that share ZrO(OH) secondary building units have been systematically examined. It is demonstrated that small pore diameters enhance short-term properties, while large pores, which are characterized by increased ion affinity, sustain long-term properties. Moreover, we demonstrated a 6 × 6 pixel artificial retina by incorporating both short-term and long-term artificial synapses with a signal-integration device. Signal summation by the signal-integration device enables attention-based information processing. The information-filterable artificial retina system developed here emulates human perception processes and holds promise in the fields of neuroprosthetics and advanced artificial intelligence.