Mitochondrial volume reduced row prediction error for calcium activity in 105 ferret spine records

We tested whether mitochondrial volume improves prediction of local calcium amplitude in a public table that links imaging and electron microscopy in ferret visual cortex. In 105 positive-volume records, a fixed linear volume rule reduced leave-one-row-out squared error by 14.3% relative to a training mean. A decreasing monotone rule reduced error by 11.1%. The linear gain remained positive after every single-row deletion. In 51 matched records, a shell-volume rule reduced error by 27.4%, while inner volume alone did not improve squared error. Added anatomy often reduced prediction quality. This study extends prior rank-association analyses with fixed prediction rules and per-row loss checks. The rows lack cell and animal labels and can share calcium neighborhoods. These are within-table diagnostics, not tests of transfer across animals or proof of a physiological mechanism. All code, inputs, provenance and predictions are included.

mitochondria · calcium imaging · prediction · structure and function · ferret visual cortex