BACKGROUND Using standard density gradient (SDG) ranges for human islet purification frequently results in islet loss and transplantation of lower islet mass. Measuring the densities of islet and acinar tissue beforehand to customize the gradient range for the actual COBE 2991 cell processor (COBE) purification is likely to maximize the recovery of islets. We developed an analytical test gradient system (ATGS) for predicting pancreatic tissue densities before COBE purification to minimize islet loss during purification. METHODS Human islets were isolated from deceased donor (n=30) and chronic pancreatitis pancreata (n=30). Pancreatic tissue densities were measured before purification by the ATGS, and the density gradient range for islet purification in a COBE was customized based on density profiles determined by the ATGS. The efficiency of custom density gradients (CDGs) to recover high islet yield was compared with predefined SDGs. RESULTS Pancreatic tissue densities from autografts were significantly higher than in allograft preparations. In allograft purifications, a higher proportion of islets were recovered using ATGS-guided CDGs (85.9%±18.0%) compared with the SDG method (69.2%±27.0%; P=0.048). Acinar contamination at 60%, 70%, and 80% cumulative islet yield for allografts was significantly lower in the CDG group. In autograft purifications, more islets were recovered with CDGs (81.9%±28.0%) than SDGs (55.8%±22.8%; P=0.03). CDGs effectively reduced islet loss by minimizing islet sedimentation in the COBE bag. CONCLUSIONS Using ATGS-guided CDGs maximizes the islet recovery for successful transplantations by reducing acinar contamination in allograft preparations and by reducing sedimentation of islets in the COBE bag in autograft preparations.