TEND-GRADS(1)

💅 tend-grads — the manual, but make it make sense

ok so this is the tend-grads page, straight out of the User Commands section of the manual (section 1, bestie, we're in the cheap seats). It's about figuring out balanced gradient directions for DWI acquisition. Let's get into it 🧚

---

✨ NAME ✨

tend-grads — calculate balanced gradient directions for DWI acquisition

---

🦋 SYNOPSIS 🦋

tend grads [-n <# dir>] [-i <grads>] [-seed <value>] [-step <step>]
[-single] [-snap <interval>] [-jitter <jitter>] [-miniter <# iters>]
[-maxiter <# iters>] [-minvelo <vel>] [-exp <exponent>] [-dp
<potential change>] [-minimprov <delta>] [-minmean <len>] [-izv
<insert>] [-o <filename>]

---

🌸 DESCRIPTION 🌸

ok so. this description is doing a whole thing. it says: composites an RGBA nrrd over a background color (or image), after doing gamma correction, then quantizes to an 8-bit image. and honestly? that is what it says. not me questioning it. 😭

anyway, more deets: the input nrrd can have more than 4 values per pixel, but only the first four are used. no cap.

if the RGBA nrrd is floating point, the values are taken at face value — like, exactly what they say on the tin. if it's fixed point, the values are interpreted as having been quantized, so that 8-bit RGBA images will act as you expect. (so that 8-bit RGBA images will act as you expect, fr fr 💖)

and when compositing with a background image? the background image you give does not have to be the same size as the input image. it will be resampled (with linear interpolation, slay) to fit. iconic behavior honestly.

---

💖 OPTIONS 💖

• -n <# dir> — desired number of diffusion gradient directions (int) default: "6"
• -i <grads> — initial gradient directions to start with, instead of default random initial directions. heads up: this overrides -n; default: ""
• -seed <value> — seed value to be used with airSrandMT() (unsigned int) default: "42"
• -step <step> — time increment in solver (double) default: "1.0"
• -single — instead of the default behavior of tracking a pair of antipodal points (appropriate for determining DWI gradients), use only single points (appropriate for who knows what). 💀
• -snap <interval> — specifies an interval between which snapshots of the point positions should be saved out. by default (not using this option), there is no such snapshot behavior (int) default: "0"
• -jitter <jitter> — amount by which to perturb points when given an input nrrd (double) default: "0.1"
• -miniter <# iters> — min number of iterations for which to run the simulation (int) default: "0"
• -maxiter <# iters> — max number of iterations for which to run the simulation (int) default: "1000000"
• -minvelo <vel> — low threshold on mean velocity of repelling points, at which point the repulsion phase of the algorithm terminates. (double) default: "0.00001"
• -exp <exponent> — the exponent n that determines the potential energy 1/r^n. (double) default: "1"
• -dp <potential change> — low threshold on fractional change of potential at which point the repulsion phase of the algorithm terminates. (double) default: "0.000000001"
• -minimprov <delta> — in the second phase of the algorithm, when stochastically balancing the sign of the gradients, the (small) improvement in length of mean gradient which triggers termination (as further improvements are unlikely). (double) default: "0.00005"
• -minmean <len> — if length of mean gradient falls below this, finish the balancing phase (double) default: "0.0001"
• -izv <insert> — adding zero vector at beginning of grads (bool) default: "false"
• -o <filename> — file to write output nrrd to (string); default: "-"

bestie the defaults are in there, the types are in there, everything's in there. don't come for me. 🫶

---

👑 SEE ALSO 👑

tend(1)

---

and that's the whole page, periodt. this was TEND-GRADS(1), dated April 2021, and it's giving main character energy from start to finish. go balance some gradients, you got this 🌈💜

---