This is the MRI Graham had on the morning of 8 October 2026: the Year-1 research scan of the BD2 (Breakthrough Discoveries for Thriving with Bipolar Disorder) Integrated Network, at UT Austin's Biomedical Imaging Center, on a Siemens MAGNETOM Prisma NX 3T. The sequence list and timings are the BD2 Appendix E "BD2 Prisma MRI Acquisition Protocol (using ABCD sequences)" with the parameters from the ABCD Siemens Prisma protocol printout (43 minutes 41 seconds of scanning in a one-hour slot). The quartet plays the whole thing in order at about one eighth of real time, every scan switching on and off dead the way the machine does. Pitches and rhythms marked (estimated) are worked out from the printed parameters — a readout tone at 1/(2 x echo spacing), the rates from TR, slice count and multiband factor — not measured in the room. The one interval in the piece is the perfect fourth Graham heard: the diffusion readout sits near F-sharp (about 725 Hz) and the resting-state BOLD readout near B (about 980 Hz), so the scan itself steps up a fourth from the diffusion to the last two BOLD runs. Under every gap, the helium cold-head pump. Nothing in it is a tune.
bar 1 (0:00.0): patient table sliding into the bore (table motor whir)
bars 2–3 (0:03.3): cold-head pump alone (helium compressor, ~1.2 Hz thump = 72/min), cello pizz. ppp
bars 4–6 (0:10.0): scan 1 — localizer, three-plane scout: GRE, TR 7.5 ms → low buzz ~133 Hz ≈ C3 (estimated), three short bursts, 9 s real
bar 7 (0:17.5): gap: pump
bar 8 (0:20.8): gap: intercom click, tech's voice (viola, two notes — "lie still")
bar 9 (0:24.2): scan 2 — T1 navigator setter, ~1 s: nav echo spacing 0.27 ms → short high chirp ~1.85 kHz ≈ A#6 (estimated; stretched to be heard)
bars 10–25 (0:26.7): scan 3 — T1 MPRAGE with navigators, 7:12 real: TR 2500 ms, TI 1070, echo spacing 8.1 ms → gritty buzz ~123 Hz ≈ B2 for ~1.4 s, quieter gap with the nav chirp, every 2.5 s = one bar; 16 of the real 173 repeats (estimated)
bar 26 (1:06.7): gap: pump
bar 27 (1:10.0): gap: intercom — "eyes open, on the cross" (fixation cross for resting state)
bars 28–30 (1:13.3): scan 4 — resting-state field map PA (spin-echo EPI), 7 s: TR 7030 ms, 60 slices, echo spacing 0.51 ms → ~980 Hz ≈ B5 clicks at ~8.5/s (estimated)
bar 31 (1:19.0): between the two field maps
bars 32–34 (1:20.8): scan 5 — resting-state field map AP: the same, reversed phase-encode direction (second violin)
bar 35 (1:26.5): gap: pump
bars 36–67 (1:29.8): scan 6 — resting-state BOLD run 1, 5:14 real: multiband 6, TR 800 ms, 60 slices, 383 volumes, echo spacing 0.51 ms → ~980 Hz ≈ B5 readout (+3rd harmonic ~2.9 kHz) chopped 12.5×/s (10 slice groups per TR, ~46 ms each, every 80 ms); one 5/8 bar = one TR; faint 1.25 Hz TR beat in the cello; the loudest scan (estimated)
bar 68 (1:55.3): gap: pump
bars 69–100 (1:58.7): scan 7 — resting-state BOLD run 2: identical to run 1 (383 volumes, 5:14 real; 32 TRs here)
bar 101 (2:24.2): gap: pump
bar 102 (2:27.5): gap: pump (the tech sets up the diffusion scans)
bars 103–104 (2:30.9): scan 8 — diffusion distortion map PA (SE-EPI), 25 s: TR 12400 ms, echo spacing 0.69 ms → ~725 Hz ≈ F#5, slower (estimated)
bar 105 (2:35.9): between the two distortion maps
bars 106–107 (2:38.4): scan 9 — diffusion distortion map AP: the same, reversed (second violin)
bar 108 (2:43.4): gap: intercom — "this one is loud"
bars 109–118 (2:46.7): scan 10 — multi-shell diffusion, 7:29 real: multiband 3, TR 4200 ms, 81 slices, b = 500/1000/2000/3000, 96 directions, echo spacing 0.69 ms → ~725 Hz ≈ F#5 readout bursts ~70 ms at 6.4/s over heavy col legno thumps from the diffusion gradients; one 7/4 bar = one TR; the thump weight lurches at every barline as the direction / b-value changes (mp–mf–f–ff = b 500→3000, estimated order)
bar 119 (3:28.7): gap: pump
bar 120 (3:32.0): gap: intercom — "halfway there"
bar 121 (3:35.4): scan 11 — T2 navigator setter, ~1 s: TR 13 ms → short chirp (estimated; stretched)
bars 122–132 (3:37.9): scan 12 — T2 SPACE with navigators, 6:35 real: TR 3200 ms, echo spacing 6.28 ms, echo train 1187 ms → low hum ~159 Hz ≈ E-flat 3 for ~1.2 s, ~2 s near-quiet with the nav chirp, every 3.2 s = one bar; softer than the T1 (estimated)
bar 133 (4:13.1): gap: pump
bars 134–136 (4:16.4): scan 13 — resting-state field map PA: as scan 4 (~980 Hz ≈ B5 clicks ~8.5/s) — THE FOURTH: the tone steps up from the diffusion F#5 to B5 (estimated from the echo spacings 0.69 vs 0.51 ms)
bar 137 (4:22.0): between the two field maps
bars 138–140 (4:23.9): scan 14 — resting-state field map AP
bar 141 (4:29.5): gap: pump
bars 142–173 (4:32.8): scan 15 — resting-state BOLD run 3: as run 1 (multiband 6, TR 800 ms, ~980 Hz chopped 12.5×/s, one 5/8 bar per TR)
bar 174 (4:58.4): gap: intercom — "last one"
bars 175–206 (5:01.7): scan 16 — resting-state BOLD run 4: as run 1; the scanning ends here (43:41 real)
bar 207 (5:27.2): the last scan stops; pump alone
bar 208 (5:30.6): patient table sliding out of the bore (table motor whir)
bars 209–210 (5:33.9): cold-head pump, then nothing