{"id":17190,"date":"2026-04-04T18:00:08","date_gmt":"2026-04-04T21:00:08","guid":{"rendered":"https:\/\/rtmedical.com.br\/aaa-eclipse-algorithm-explained\/"},"modified":"2026-04-04T19:57:56","modified_gmt":"2026-04-04T22:57:56","slug":"aaa-eclipse-algorithm-explained","status":"publish","type":"post","link":"https:\/\/rtmedical.com.br\/en\/aaa-eclipse-algorithm-explained\/","title":{"rendered":"AAA Explained: Beamlets, Anisotropic Kernels, and Heterogeneity Correction in Eclipse"},"content":{"rendered":"<p><span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> stayed at the center of clinical photon planning long enough to become a default assumption in many departments. That is exactly why it is often described too loosely. People call it \u201cbetter than pencil beam\u201d and \u201clighter than explicit transport,\u201d which is true in a broad market sense but not very useful if the goal is to understand how the algorithm actually behaves in a plan.<\/p>\n<p>Inside <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Eclipse<\/span>, the <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Anisotropic Analytical Algorithm<\/span> is more interesting as engineering than as branding. It was designed to pull much more physics into routine photon dose calculation without pushing the workflow into the computational cost of explicit transport solvers. That balance is what made it clinically important, and it is also what explains its limitations.<\/p>\n<p>To read <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> properly, it helps to break the problem into three layers:<\/p>\n<ol>\n<li>how the beam is modeled before entering the patient;<\/li>\n<li>how the treatment field is decomposed into <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">beamlets<\/span>;<\/li>\n<li>how longitudinal and lateral energy deposition are handled in heterogeneous tissue.<\/li>\n<\/ol>\n<p>The point is not simply to repeat that <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> \u201caccounts for heterogeneity.\u201d The useful question is what it does, where it remains very strong, and why the Varian documentation itself recommends caution in specific scenarios.<\/p>\n<div class=\"toc\">\n<h2>In this Article<\/h2>\n<ul>\n<li><a href=\"#what-aaa-actually-is\">1. What AAA actually is<\/a><\/li>\n<li><a href=\"#the-algorithm-starts-before-the-patient\">2. The algorithm starts before the patient<\/a><\/li>\n<li><a href=\"#beamlets-the-working-unit-of-aaa\">3. Beamlets: the working unit of AAA<\/a><\/li>\n<li><a href=\"#the-central-expression-in-aaa\">4. The central expression in AAA<\/a><\/li>\n<li><a href=\"#radiological-depth-where-heterogeneity-really-enters\">5. Radiological depth: where heterogeneity really enters<\/a><\/li>\n<li><a href=\"#the-lateral-kernel-and-what-anisotropic-really-means\">6. The lateral kernel and what \u201canisotropic\u201d really means<\/a><\/li>\n<li><a href=\"#why-the-algorithm-is-built-around-energy-not-just-dose\">7. Why the algorithm is built around energy, not just dose<\/a><\/li>\n<li><a href=\"#separate-beam-components-matter-clinically\">8. Separate beam components matter clinically<\/a><\/li>\n<li><a href=\"#where-aaa-remains-strong-clinically\">9. Where AAA remains strong clinically<\/a><\/li>\n<li><a href=\"#where-varian-itself-recommends-caution\">10. Where Varian itself recommends caution<\/a><\/li>\n<li><a href=\"#grid-resolution-is-not-a-side-issue\">11. Grid resolution is not a side issue<\/a><\/li>\n<li><a href=\"#aaa-is-not-a-lighter-acuros\">12. AAA is not a lighter Acuros<\/a><\/li>\n<li><a href=\"#when-aaa-is-still-the-right-choice\">13. When AAA is still the right choice<\/a><\/li>\n<\/ul>\n<\/div>\n<h2 id=\"what-aaa-actually-is\">What AAA actually is<\/h2>\n<p>In the <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Eclipse 17.0<\/span> documentation, <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> is described as a <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">3D pencil beam convolution\/superposition<\/span> algorithm. That short definition matters more than it seems.<\/p>\n<p><span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> is not part of the explicit transport family represented by engines such as <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Acuros XB<\/span>. It still belongs to the <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">pencil beam<\/span> lineage, but to a much more advanced branch of it. Instead of relying on simple path-length corrections to mimic scatter and heterogeneity, it combines:<\/p>\n<ul>\n<li>a source model split into distinct beam components;<\/li>\n<li><span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Monte Carlo<\/span>-derived kernels;<\/li>\n<li>field decomposition into finite <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">beamlets<\/span>;<\/li>\n<li>radiological scaling and anisotropic handling of lateral spread.<\/li>\n<\/ul>\n<p>That is why <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> became such an important clinical engine. It offers a substantial jump over classical <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">pencil beam<\/span> approaches without asking the user to pay the full computational price of explicit transport.<\/p>\n<h2 id=\"the-algorithm-starts-before-the-patient\">The algorithm starts before the patient<\/h2>\n<p>One common mistake in explaining <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> is to start directly with dose deposition in tissue, as if the physics of the problem began at the skin surface. It does not. Before any convolution happens, the clinical beam itself has to be modeled.<\/p>\n<p>In <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Eclipse<\/span>, <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> and <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Acuros XB<\/span> share the same <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">photon beam source model<\/span>. That means both algorithms start from a common description of head scatter and incident fluence built from configuration data and measurements. In practice, this source model separates at least four relevant contributions:<\/p>\n<ul>\n<li>the primary photon source;<\/li>\n<li>an extra-focal photon source;<\/li>\n<li>electron contamination;<\/li>\n<li>scatter associated with modifiers such as wedges.<\/li>\n<\/ul>\n<p>This separation is not cosmetic. Primary photons, extra-focal photons, and contaminating electrons do not have the same spatial or spectral behavior. Modeling them independently improves the field description before depth-dose and lateral spread are even considered.<\/p>\n<p>That is one reason why <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> was such a visible improvement over simpler engines. The gain does not come only from \u201cmore corrections.\u201d It comes from starting with a more physical beam model.<\/p>\n<h2 id=\"beamlets-the-working-unit-of-aaa\">Beamlets: the working unit of AAA<\/h2>\n<p>After the clinical fluence is modeled, the open or modulated field is decomposed into finite <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">beamlets<\/span>. The classic Varian figure, extracted from the local documentation set, helps make that step concrete:<\/p>\n<figure class=\"wp-block-image size-large\"><img alt=\"Beamlet geometry and coordinate systems used by AAA\" class=\"alignright lazyload\" decoding=\"async\" data-src=\"https:\/\/rtmedical.com.br\/wp-content\/uploads\/2026\/04\/eclipse_aaa_beamlet-086.png\" src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" style=\"--smush-placeholder-width: 1275px; --smush-placeholder-aspect-ratio: 1275\/1650;\" \/><\/figure>\n<p>Each <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">beamlet<\/span> acts as a local calculation unit. Instead of treating the entire field as a single object, <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> sums the contributions of many small beam elements. That keeps the logic of the <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">pencil beam<\/span> family while making it much more powerful once it is combined with <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Monte Carlo<\/span>-derived kernels and anisotropic heterogeneity scaling.<\/p>\n<p>In the Varian manual, the lateral width of a <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">beamlet<\/span> is tied to the selected calculation resolution in the isocenter plane. That detail is clinically important. Grid resolution is not only a display choice. It also affects how the field is discretized internally.<\/p>\n<p>Each <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">beamlet<\/span> carries:<\/p>\n<ul>\n<li>local fluence;<\/li>\n<li>spectral weighting;<\/li>\n<li>geometric position;<\/li>\n<li>the parameters required for longitudinal deposition and lateral scatter.<\/li>\n<\/ul>\n<p>Dose at a given point is then obtained by superposing the contributions of all relevant <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">beamlets<\/span>. That architecture helps explain why <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> remained fast enough for routine planning while still handling fairly complex fields.<\/p>\n<h2 id=\"the-central-expression-in-aaa\">The central expression in AAA<\/h2>\n<p>The Varian manual presents the energy associated with a <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">beamlet<\/span> using an expression of the form:<\/p>\n<div class=\"dose-formula\" style=\"margin:1.25rem 0;padding:0.9rem 1.1rem;border-left:4px solid #0f4c81;background:#f5f8fb;color:#102a43;font-family:'Cambria Math','STIX Two Math','Times New Roman',serif;font-size:1.05rem;line-height:1.7;overflow-x:auto;\"><span style=\"white-space:nowrap;\">E<sub>ph,\u03b2<\/sub>(X,Y,Z)=\u03a6<sub>\u03b2<\/sub> \u00d7 I<sub>\u03b2<\/sub>(z,\u03c1)\u00d7 K<sub>\u03b2<\/sub>(X,Y,Z)<\/span><\/div>\n<p>This compact expression is one of the best ways to understand the algorithm because it condenses its core logic.<\/p>\n<h3>Fluence term: \u03a6\u03b2<\/h3>\n<p>This term represents how much radiation that <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">beamlet<\/span> carries. Source shaping, collimation, and beam modulation all influence it.<\/p>\n<h3>Longitudinal term: I\u03b2(z,\u03c1)<\/h3>\n<p>This term describes how energy is deposited along depth for that <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">beamlet<\/span>, taking tissue density into account. It captures attenuation and the depth-dependent behavior of the beam.<\/p>\n<h3>Lateral kernel: K\u03b2(X,Y,Z)<\/h3>\n<p>This kernel distributes energy laterally around the beamlet path. That is one of the pieces that makes <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> much more capable than simplified <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">pencil beam<\/span> models.<\/p>\n<p>The key point is that <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> does not apply one global correction to the entire field. It combines a local fluence description, a local longitudinal deposition term, and a local lateral spread term for each <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">beamlet<\/span>.<\/p>\n<h2 id=\"radiological-depth-where-heterogeneity-really-enters\">Radiological depth: where heterogeneity really enters<\/h2>\n<p>One of the central ideas in <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> is the use of radiological depth rather than geometric depth alone:<\/p>\n<div class=\"dose-formula\" style=\"margin:1.25rem 0;padding:0.9rem 1.1rem;border-left:4px solid #0f4c81;background:#f5f8fb;color:#102a43;font-family:'Cambria Math','STIX Two Math','Times New Roman',serif;font-size:1.05rem;line-height:1.7;overflow-x:auto;\"><span style=\"white-space:nowrap;\">z&#8217;=\u222b<sub>0<\/sub><sup>Z<\/sup> (\u03c1(0,0,t))\/(\u03c1<sub>water<\/sub>),dt<\/span><\/div>\n<p>This equation looks simple, but it carries an important physical interpretation. The algorithm attempts to rewrite a heterogeneous beam path as an equivalent water path, weighted by relative density.<\/p>\n<p>That is exactly what allows <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> to improve substantially over older methods. The beam does not just \u201csee\u201d how many centimeters it has traveled. It sees how many radiological centimeters it has crossed.<\/p>\n<p>But this is also where a conceptual limitation appears. Radiological depth handles part of the heterogeneity problem mainly along the main beam direction. It does not, by itself, resolve all the lateral transport effects associated with loss of charged-particle equilibrium or complex interface physics. That is why <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> improves strongly over classical <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">pencil beam<\/span> methods without becoming explicit transport.<\/p>\n<h2 id=\"the-lateral-kernel-and-what-anisotropic-really-means\">The lateral kernel and what \u201canisotropic\u201d really means<\/h2>\n<p>The word <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">anisotropic<\/span> in <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> is not decoration. It points to one of the algorithm\u2019s most important upgrades.<\/p>\n<p>Classical <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">pencil beam<\/span> approaches often assume a fairly simple lateral dose-spread behavior. <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> goes further by allowing the lateral energy redistribution to respond to the medium in a direction-dependent way. In the Varian documentation, the lateral kernel is represented through a sum of terms with depth-dependent coefficients:<\/p>\n<div class=\"dose-formula\" style=\"margin:1.25rem 0;padding:0.9rem 1.1rem;border-left:4px solid #0f4c81;background:#f5f8fb;color:#102a43;font-family:'Cambria Math','STIX Two Math','Times New Roman',serif;font-size:1.05rem;line-height:1.7;overflow-x:auto;\"><span style=\"white-space:nowrap;\">K<sub>\u03b2<\/sub>(x,y,z)=\u2211 c<sub>k<\/sub>(z&#8217;)(1)\/(r)e<sup>-\u03bc<sub>k<\/sub> r<\/sup><\/span><\/div>\n<p>You do not need to read this as a full derivation to understand its significance. The important point is that lateral spread is not treated as a fixed blur. It changes with radiological depth and with the beamlet context.<\/p>\n<p>That is a major reason why <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> behaves much better than simple <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">pencil beam<\/span> engines near density changes and in fields where lateral scatter matters.<\/p>\n<h2 id=\"why-the-algorithm-is-built-around-energy-not-just-dose\">Why the algorithm is built around energy, not just dose<\/h2>\n<p>Another useful reading of <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> is that it is fundamentally an energy-transport approximation before it becomes a displayed dose map. The kernels are derived from <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Monte Carlo<\/span> data and then scaled to clinical conditions. That matters because it anchors the algorithm in a physically informed representation of how secondary particles spread energy around the primary beam path.<\/p>\n<p>This is also why <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> should not be dismissed as \u201cold-fashioned\u201d just because newer transport engines exist. Within its own family, it brought a real increase in physical realism.<\/p>\n<h2 id=\"separate-beam-components-matter-clinically\">Separate beam components matter clinically<\/h2>\n<p>The Varian documentation also emphasizes that <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> handles beam components separately:<\/p>\n<ul>\n<li>primary photons;<\/li>\n<li>extra-focal photons;<\/li>\n<li>wedge-related scatter;<\/li>\n<li>electron contamination.<\/li>\n<\/ul>\n<p>That separation improves:<\/p>\n<ul>\n<li>surface dose behavior;<\/li>\n<li>penumbra description;<\/li>\n<li>near-edge dose;<\/li>\n<li>the handling of modified fields.<\/li>\n<\/ul>\n<p>For example, electron contamination is modeled with its own depth dependence and lateral parameters. This is one reason why <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> behaves much better than simplified <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">pencil beam<\/span> models in shallow regions where surface and build-up behavior matter.<\/p>\n<p>At the same time, this reminds us of something every physicist already knows but many comparisons ignore: a strong dose algorithm still depends on a strong <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">beam model<\/span>. If source settings, contamination parameters, or beam data are poor, the label <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> will not rescue the calculation.<\/p>\n<h2 id=\"where-aaa-remains-strong-clinically\">Where AAA remains strong clinically<\/h2>\n<p><span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> is still a very competent algorithm across a large part of the photon-planning routine. Its strength comes from the combination of:<\/p>\n<ul>\n<li>a solid source model;<\/li>\n<li>anisotropic lateral-spread handling;<\/li>\n<li><span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Monte Carlo<\/span>-derived kernels;<\/li>\n<li>calculation times compatible with optimization and routine planning.<\/li>\n<\/ul>\n<p>It generally performs very well in:<\/p>\n<ul>\n<li>conventional <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">3D-CRT<\/span>;<\/li>\n<li>a large share of <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">IMRT<\/span> and <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">VMAT<\/span> workflows;<\/li>\n<li>moderate heterogeneity scenarios;<\/li>\n<li>cases where the gain over classical <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">pencil beam<\/span> matters much more than the incremental gain from explicit transport.<\/li>\n<\/ul>\n<p>This is the real reason <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> occupied such stable clinical ground for so long. It delivers a level of physical fidelity that is already decisive for many practical cases without imposing the heavier operational cost of transport solvers.<\/p>\n<h2 id=\"where-varian-itself-recommends-caution\">Where Varian itself recommends caution<\/h2>\n<p>The healthiest way to read a commercial algorithm is not to look only at what it does well. It is also to look carefully at what the manufacturer explicitly flags as a limitation.<\/p>\n<p>In the <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Eclipse<\/span> guide, Varian calls attention to known <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> limitations, especially in lung. There is a direct warning that when the beam crosses lung and then re-enters soft tissue, the algorithm tends to overestimate dose at the exit interface. The effect becomes more pronounced as the interface lies deeper in the beam path.<\/p>\n<p>That statement deserves to be read slowly. It is not a random failure. It is a structural limitation tied to the kind of approximation the algorithm uses to represent transport and scatter in strongly heterogeneous media.<\/p>\n<p>The manual also mentions performance differences in scenarios involving:<\/p>\n<ul>\n<li>specific field sizes and energies in lung;<\/li>\n<li>static MLC-defined fields under certain conditions;<\/li>\n<li>physical wedges;<\/li>\n<li>supports and high-density structures.<\/li>\n<\/ul>\n<p>This is exactly the kind of guidance that should shape clinical judgment. An algorithm does not have to be rejected for its weaknesses to be taken seriously. What makes no sense is to use a scaled-kernel engine in a highly critical heterogeneity case and then act surprised when a more physical motor shifts the dose interpretation.<\/p>\n<h2 id=\"grid-resolution-is-not-a-side-issue\">Grid resolution is not a side issue<\/h2>\n<p>The <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Eclipse<\/span> guide also highlights something often treated as merely operational: grid resolution.<\/p>\n<p>In <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span>, the resolution typically ranges from <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">1<\/span> to <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">5 mm<\/span>, and dose is first computed in a divergent matrix before being converted to the output grid. The relation between selected resolution, pixel spacing, and slice separation affects how the final dose is sampled.<\/p>\n<p>That has real consequences. In high-gradient regions or in small fields, apparent algorithm behavior can change substantially with grid choice. In other words, part of what is sometimes reported as \u201cdifference between algorithms\u201d is really a difference between discretizations of the same physical problem.<\/p>\n<p>Methodologically, the implication is straightforward: comparing <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> with any serious transport engine without controlling for grid, slice spacing, and sampling parameters is not a clean comparison.<\/p>\n<h2 id=\"aaa-is-not-a-lighter-acuros\">AAA is not a lighter Acuros<\/h2>\n<p>A recurring mistake is to think of <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> as if it were simply a lighter version of <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Acuros<\/span>. It is not.<\/p>\n<p>The two algorithms share part of the beam modeling, but they belong to different conceptual families:<\/p>\n<table>\n<thead>\n<tr>\n<th>Aspect<\/th>\n<th>AAA<\/th>\n<th>Acuros XB<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Family<\/td>\n<td>Beamlet-based convolution\/superposition<\/td>\n<td>Numerical solution of the LBTE<\/td>\n<\/tr>\n<tr>\n<td>Main physical basis<\/td>\n<td><span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Monte Carlo<\/span>-derived kernels and radiological scaling<\/td>\n<td>Explicit discretized transport in space, energy, and angle<\/td>\n<\/tr>\n<tr>\n<td>Most central image quantity<\/td>\n<td>Electron density<\/td>\n<td>Mass density and material composition<\/td>\n<\/tr>\n<tr>\n<td>Classical weakness<\/td>\n<td>Strong heterogeneous interfaces, especially lung<\/td>\n<td>Sensitivity to discretization and material mapping<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>That conceptual distinction matters because it changes both case interpretation and validation strategy.<\/p>\n<h2 id=\"when-aaa-is-still-the-right-choice\">When AAA is still the right choice<\/h2>\n<p>There is a current temptation to treat older engines as if they were automatically inferior in every setting. That is poor reading.<\/p>\n<p><span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> still makes very good clinical sense when:<\/p>\n<ul>\n<li>the case is not dominated by extreme heterogeneity;<\/li>\n<li>the workflow needs speed without giving up solid physics;<\/li>\n<li>the institution knows its <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">beam model<\/span> well;<\/li>\n<li>comparison with more physical engines does not materially change the clinical decision.<\/li>\n<\/ul>\n<p>That is why <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> remains relevant. It is not an outdated engine that happened to survive. It is a carefully built compromise between source modeling, <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">Monte Carlo<\/span>-derived kernels, clinical speed, and a far more serious treatment of heterogeneity than classical <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">pencil beam<\/span> ever offered.<\/p>\n<p>That is also what explains its limits. In lung, at strong interfaces, and in scenarios where lateral transport dominates the problem, <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> still belongs to the world of scaled kernels. It can go very far inside that family, but it does not become explicit transport by approximation.<\/p>\n<p>Reading the algorithm this way is more useful than repeating labels. It helps answer the practical questions that matter: when <span class=\"tg263-code\" style=\"display:inline;padding:0.08em 0.45em;margin:0 0.08em;border:1px solid rgba(18,42,66,0.14);border-radius:6px;background:#eef3f8;color:#17324a;font-family:Menlo,Monaco,Consolas,'Courier New',monospace;font-size:0.92em;font-weight:600;line-height:1.2;white-space:nowrap;vertical-align:baseline;\">AAA<\/span> is still the right engine, when recalculation with a more physical solver is worth the effort, and which discrepancies should be read as expected limitations rather than as a surprise from the TPS.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>A technical and clinical reading of AAA: source model, beamlets, energy convolution, heterogeneity, and known lung limits.<\/p>\n","protected":false},"author":0,"featured_media":17165,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"om_disable_all_campaigns":false,"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"ngg_post_thumbnail":0,"fifu_image_url":"","fifu_image_alt":"","footnotes":""},"categories":[98,60],"tags":[],"class_list":{"0":"post-17190","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-radioterapia","8":"category-software"},"aioseo_notices":[],"rt_seo":{"title":"AAA Explained: Beamlets, Anisotropic Kernels, and Heterog...","description":"Understand how Eclipse AAA uses beamlets, anisotropic kernels, and radiological depth to calculate dose in heterogeneous media.","canonical":"https:\/\/rtmedical.com.br\/aaa-eclipse-algorithm-explained\/","og_image":"https:\/\/rtmedical.com.br\/wp-content\/uploads\/2026\/04\/eclipse_aaa_beamlet-086.png","robots":"index,follow","schema_type":"Article","include_in_llms":true,"llms_label":"Guia t\u00e9cnico","llms_summary":"A technical and clinical reading of AAA: source model, beamlets, energy convolution, heterogeneity, and known lung limits.","faq_items":[],"video":[],"gtin":"","mpn":"","brand":"","aggregate_rating":[]},"_links":{"self":[{"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/17190\/"}],"collection":[{"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/"}],"about":[{"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/types\/post\/"}],"replies":[{"embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/comments\/?post=17190"}],"version-history":[{"count":3,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/17190\/revisions\/"}],"predecessor-version":[{"id":17276,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/posts\/17190\/revisions\/17276\/"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/media\/17165\/"}],"wp:attachment":[{"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/media\/?parent=17190"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/categories\/?post=17190"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/rtmedical.com.br\/en\/wp-json\/wp\/v2\/tags\/?post=17190"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}