Skip to main content
Search

Universidad Complutense de Madrid

2 related articles

Subscribe to updates
  1. bioRxiv

    INME, a circuit using only four cardinal-direction cells, reproduces eleven motion phenomena while million-parameter optical-flow networks miss most

    Motion direction estimation has lacked a unified model spanning species and brain regions, since existing models tend to fit one animal class, one brain area or one phenomenon. INME uses only cells selective to the four cardinal directions, embedding dendritic nonlinearity and a back-propagating action potential in one circuit with a post-summation readout stage. Across eleven synthetic-stimulus experiments it qualitatively reproduces phenomena from contrast saturation and reverse-Phi to the breaking of time-reversal symmetry, while five deep optical-flow networks with millions of parameters reproduce only a few of them.

  2. bioRxiv

    On a synthetic pulmonary-artery benchmark, biomarker supervision cut 4D flow MRI pulsatility-index error from 17.12% to 12.76%

    The study trained a 3D residual channel attention network (RCAN) for 2x super-resolution of 4D flow MRI on a synthetic pulmonary-artery benchmark and found that adding biomarker supervision reduced pulsatility-index (PI) error under both regularization conditions (in the original primary comparison, 12.76 +/- 2.34% vs. 17.12 +/- 1.23%; paired difference -4.35 percentage points [95% CI -5.58, -3.09]; p = 0.0015), while PI improvement was not mirrored by uniformly improved reconstruction metrics (PSNR changes were not significant; SSIM decreased by 0.009 and 0.007) and only adjacent-frame temporal-difference, not divergence, regularization was associated with reproducible PI deterioration.

You’re all caught up