no idea. I thought it is labeled wrong, the left should be "WFS = 3 pixels = 1.5mm" instead. also the conclusion at 24:20 should be "higher bandwith is both giving us a decrease in signal-to-noise as well as decreasing the amount of chemical shift effects"

The numbers appear to be BW of 57 kHz in the right image, and about 19 kHz (19.3 I believe) in the left image. If we divide 57 kHz into a 256-pixel width in the frequency-encoding direction, that means each pixel is 57 kHz / 256 = 222 Hz wide. Given a chem shift of 224 Hz at 1.5T, that results in a chem shift spatially of almost exactly one pixel, just as is reported. Conversely, with a 19 kHz bandwidth, each pixel is 19 kHZ / 256 = 74 Hz wide, so a 224 Hz shift translates to 3 pixels spatially. Also, I agree with spdn. The left image should be labelled WFS of 3 pixels, 1.5mm. And he misspoke at 24:20, the higher bandwidth in the right image results in a tradeoff, decreased chem shift effect but reduced SNR.

yes

Yes, in his diagram of the water/olive oil example he misdrew the chem shift in the opposite direction. Fat signal should be shifted against the frequency encoding direction, so really what should happen in that diagram is a bright stripe on the rightward border of the water, and a dark stripe on the rightward border of the oil. Similarly, in his example image of the psoas muscles, the frequency encoding direction is in the opposite direction of how he labelled the images. We see fat signal is shifted to the right on the image (patient's left), which means the frequency encoding gradient is from right-to-left on the image (or, left-to-right for the patient).

thanks man, i wasted 15 minutes because of that