One midget bipolar source had over half of the input counts in 290 of 341 selected ganglion cells in the HFseg1 draft human retina graph

We tested whether midget bipolar cells provide a dominant input in one public human retinal graph. We used HFseg1 table version 20250805. Of 532 listed midget targets, 386 had a listed same-polarity midget input. A coverage rule selected 341 with at least 20 input counts from listed bipolar cells. One same-polarity midget source had over half of all incoming counts in 290 selected targets (85.0%). The total included inputs from unlisted segments. The median share was 62.7% in ON cells and 62.3% in OFF cells. Mutually preferred pairs carried 91.6% and 90.9% of same-polarity midget counts. In a linear model with count weights and equal source variance, the main source explained a median 97.7% and 96.6% of response variance when sources were independent. In separate target models, all other included inputs shared one signal. These values fell to 76.1% and 75.9%. The main source stayed independent of the other inputs. These are model results. The graph alone cannot fix function. All selected targets were marked as not yet proofread. The subset excludes 191 listed targets, and many source identities remain unlisted. This study supports a dominant assigned channel in the tested graph. It does not prove a single-cone receptive field. We preserve derived tables, figures, source hashes, independent checks, and repeat code.

human retina · fovea · midget bipolar cells · ganglion cells · synaptic counts · signal covariance · connectome coverage