⬡ Proton Beam Therapy — Hepatic MFO Inverse Planner

65 y · 178 cm · 12% BF · ≈76 kg · BMI 24 Slice T12–L1 · porta hepatis RBE ≡ 1.10 · Bortfeld range σ ±2.5% v4-restored · fully offline

1 · Why protons — the Bragg peak / zero exit dose

A proton deposits little dose on entry, a sharp maximum where it stops, and essentially zero exit dose. Photons attenuate exponentially and exit through the patient. Range follows the Bortfeld (1997) power law:

R(E₀) = α·E₀p,   α = 0.0022 cm·MeV−p, p = 1.77

2 · Tissue, density, WEPL

Protons don't care what a tissue is — only how much it slows them. Stopping-power ratio maps geometry to water-equivalent depth:

WEPL(s) = ∫₀s SPR(r(s′)) ds′

Adipose (ρ≈0.92, SPR≈0.94) advances the peak ≈0.06 mm/mm → the scheduler shifts that layer's energy down; muscle (ρ≈1.06) retards it → energy up. At 12% BF this patient has ≈9 mm subcutaneous adipose and a ≈13 mm rectus wall, but a ≈26 mm paraspinal band on posterior fields — every ray gets its own energy-layer schedule, recomputed each optimizer pass.

3 · Biology: RBE is tissue-invariant

DRBE = 1.1 × Dphys   ∀ soft tissue

Fat and muscle respond identically; density differences are purely kinematic (range modulation), never biological sensitivity. All doses here are Gy(RBE).

4 · Inverse planning (MFO)

minw≥0   α·Σi∈T₁∪T₂(Di−Drx)² + β·Σj∈liverDj²   s.t.   Dk ≤ Dlim ∀k∈PV

Stage 1 — projected AdaGrad on per-spot fluence weights; PV constraint via an augmented-Lagrangian penalty that escalates while violated. Stage 2 — a coordinate search edits the computed scanning plan (energy layers ±1.5 MeV, gantry ±4°), keeping only edits that lower the objective; every edit is logged.

5 · Why the shallowest path is rejected

The anterior path has least WEPL, but its distal fall-off (80→20% over ~3–5 mm) then points straight at the portal vein. Under ±2.5% range uncertainty a 1σ overshoot converts that fall-off into near-full dose inside the vessel. The optimizer prefers geometrically longer posterior/right-oblique paths with stable density whose distal edges terminate inside T₁/T₂.

Multi-field dilution: each of 3 fields carries ~⅓ of target dose, so a ~25% entrance plateau deposits ≲8% of Drx along its own channel; the 100% cloud exists only where all SOBPs intersect — on the targets.

6 · Patient-derived slice model (T12–L1)

StructureModel valueBasis
Torso cross-section30.0 × 21.2 cm ellipselean male, waist ≈84 cm
Subcutaneous adipose≈9 mm, SPR 0.9412% BF
Abdominal wall / paraspinal≈13 mm / ≈26 mm, SPR 1.06athletic musculature
Liverright-lobe-dominant, SPR 1.05≈1500 mL organ
Portal vein (OAR)Ø 14 mm @ porta, Dmax ≤ 12 Gy(RBE)main trunk
IVC / aortaØ 18 / 16 mmstandard
L1 + ribsSPR 1.62cortical/trabecular
Stomach gasSPR ≈ 0.001large range effect
T₁ (seg VIII) / T₂ (seg VII)Ø 9 mm / Ø 8 mmright-lobe lesions

Default fields 170° / 250° / 305° (0° = anterior AP, increasing toward patient-left): posterior + right obliques that avoid stomach gas, cord and the PV distal-edge alignment. Not a clinical TPS.

Plan controls

Optimization log

Axial slice · ray marching · isodose  (radiological view: image-left = patient-right)

adipose 0.94 muscle 1.06 liver 1.05 bone portal vein (OAR) gas T₁/T₂
isodose: 20% 50% 80% 95% 107% · beam colour: blue entry → red Bragg peak → stop

Depth–dose along beam

adipose interval (schedule ↓E) muscle (↑E) T₁/T₂ WEPL windows ┄ 6 MV photons (exit tail)

Plan metrics

StructureDmeanDmaxMetricValueStatus
T₁D95
T₂D95
Portal veinDmax≤lim
Uninvolved liverV₃₀

Scanning schedule — energy layers × spots

Cell brightness = weight wk · ▼ net-adipose ray (↓E) · ▲ net-muscle (↑E) · amber = edited by Stage-2 refiner

Cumulative DVH

Convergence

3D beam geometry — transparent anatomy, WEPL segmentation, distal stop (software renderer, no CDN)

drag = orbit · wheel = zoom · beam segment colour = tissue traversed (tan = adipose ↓E · red = muscle ↑E · cyan = liver)