RMS errors from 1 µm and 10 µm grids are larger than endpoint deficits in 12 mouse traces.
Across 12 fixed mouse neuron traces with known missed crossings, the root mean square (RMS) error from uniform offsets of 1 µm and 10 µm radius grids exceeded the continuous endpoint-area deficit in all 24 trace-spacing pairs. The continuous deficit was 0.0002198854% to 0.0072634590% of reference area. RMS grid error was 0.08005784% to 0.30086874% at 1 µm spacing and 0.7584506% to 2.8898420% at 10 µm spacing. We integrated over every offset interval; offsets were not randomly sampled. The result concerns whole-range area for the stored straight segments, the stated soma center, and this fixed sample. It does not estimate a population error rate, biological uncertainty, tissue accuracy, or effects on cell classification. Endpoint grid RMS was lower than reference grid RMS in 18 of 24 pairs because opposing errors can cancel. The analysis does not directly represent finite-window trapezoidal integration.
Sholl analysis · neurons · morphology · grid offset · root mean square error · numerical integration · mouse · SWC
References
4- Endpoint counts are too small at some radii in all 12 tested mouse neuron traces. sci:sha256:792fac9fc3483c4d4e4e946aeee127b1eb8db318609a15ec95e2541929665959
[Legacy relation: reproduces] New source-file calculations reproduce all 78 missed-radius interval bounds and all 4,968 source-grid count pairs. This relation does not include the NeuroM or MorphIO library execution.
- usesSegment division at the nearest point gives the reference Sholl counts in 5,046 tests on 12 mouse neuron traces. sci:sha256:0141fa2c3dd19d57e0bae626edf8d7b8add56b257ad3ef8c84941187efbf5c3b
The nearest-point division supplies monotone radial intervals for continuous area and complete grid-offset integration. New source calculations also agree with all 78 published split parameters and coordinates.
- doi:10.1038/s41593-019-0417-0
- doi:10.1016/j.jneumeth.2014.01.017
Cited by
4 publicationsPublic catalog entries on Sciscape.
- A new calculation gives 24 grid RMS errors within 1e-40 count·µm of the reference values for 12 mouse neuron traces.
[Legacy relation: reproduces] A separate covariance calculation gives the 24 reference grid RMS errors and 24 endpoint RMS errors from the same 12 traces. It also gives 36 area and deficit values. All 84 values differ from the target by at most 1e-40 count·µm. The tested radius spacings are 1 and 10 µm.
- uses10 µm grids give a 67.98% median relative RMS error for 1 µm center movements in 12 mouse traces with soma connections.
We used its complete uniform-offset integration method. It enabled measurement of grid error for the absolute difference between paired center profiles. We did not use separate area errors as the paired difference error.
- Absolute Sholl area changes are less than half the L1 profile differences in 144 center movements on 12 mouse traces.
[Legacy relation: reproduces] A new 50-digit calculation agrees with all 12 baseline areas and all 12 endpoint deficits within 1e-40 count·µm. This relation applies to these 24 baseline values. It does not claim reproduction of the radius-grid RMS results.
- Grid RMS errors exceed endpoint deficits in 24 checks on 12 mouse traces
[Legacy relation: reproduces] New covariance code recovered all 48 RMS values, 36 continuous area values, and 48 zero-offset grid values on the same 12 traces. All 24 grid RMS values are larger than the endpoint deficits.