視頻課程之配套課件期中答疑

1、5/20/2015Xiangsheng (Jason) Yan, MS, DABR, MM.(yanxxs at )Radiation Oncology & CyberKnife, U of Arizona Cancer Center at St. Josephs Hospital & MCGammaKnife, Barrow Neurological Institute.5. 其他前期問題答疑4. 逆向計(jì)劃基本原理及目標(biāo)方程3.基本原理2. Pinnacle/Eclipse TPS commissioning 實(shí)操1. Breast Cancer Treatment Plan

2、ning 實(shí)操期中答疑：第一課至第十課問題匯總：1）MR-Linac, CyberKnife 會在第 13 講專門涉及。(5.3)2）各種機(jī)器優(yōu)缺點(diǎn)對照表，在第 13 講會專門做一張這樣的表格。3）具體例子講解透，或一個(gè)例子講幾個(gè)治療方案(乳)4）精講細(xì)講 case (乳)5）6）none 7）none，TPS Commissioning (Beam ming)8）IMRT 優(yōu)化過程()，(Emami,QUANTEC,RTOG Wiki)，再詳細(xì)一些(5.2)？9）提供學(xué)習(xí)資料、途徑、和。IMRT Qsing rate (5.1).(5.4).10）noneplanHow to compare

3、 and chose a goodIMRT 與 VAMT 差別優(yōu)化的參數(shù)設(shè)置逆向優(yōu)化國內(nèi)外醫(yī)學(xué)物理發(fā)展的差距1)2)3)4)5)6)Ant/Post, Sup/Inf, Left/Right. Supine position, Prone position.PDD, Profile, GTV à CTV (IM) à ITV à PTV (SM)why need CT to do plan: CT #, e- density, massSAD setup, SSD setup, DRR, BEV3D-CRT,IMRT, VMAT, SRS/SRT, SBRT(SA

4、BR).CI, HI, GI. Tolerance Table (Emmami à QUANTECdensity左邊鏈接à RTOG)1)2)Curative, Adjuvant, palliative3D-CRT 45%, IMRT/VMAT 45%, SRS 5-8%, SBRT/SABR 5-8%,others 10%, e.g.brachy (HDR& seed implant), TBI, TESI, adaptive, IORT, etc.基本概念二：基本概念一：一. Breast Cancer Treatment Planning 實(shí)操1.1）Gene

5、ral Points:- mastectomy, or lumpectomy + RT, depends on risk factors and staging. RT begin 2-4 weeks after surgery.- neo-adjuvant chemotherapy can convert 20-30% pts from mastectomy to lumpectomy eligible.- overall survival rate: 50-60% at 5 yrs, 30-40% at 10 yrs.- RT principles:*. 45-50 Gy at 1.8-2

6、.0 Gy/fx to whole breast (lumpectomy) or whole chest wall (mastectomy).*. Boost to tumor bed to 60-66 Gy, using electron if not deep. (e.g. 50Gy/25fx to breast & supra + 10Gy/5fx boost)*. most cases need same dose to supraclavicular fossa.*. Bolus needed (mastectomy), every other day, or until m

7、oist desquamation occurs.1.2) 45-50 Gy at 1.8-2.0 Gy/fx to whole breast, boost to tumor bed- two tangential beams (iso on the match line at posterior side)- to get max skin dose, use 6X as much as possible. If separation >= 20cm, increase more portion of 18X beams.- dose to OARs: left/right lung:

8、 V20 < 30%35%, heart: Dmean < 710Gy, V25 < 1015%.- tumor bed + 1.5cm2.0cm margin = boost area. Mostly use electron is not deep. May use photon if tumor bed deep.Boost field use electrons mostly, may use photon as well.Breast tangential fields: wedge + Field-in-Field (FinF)most cases need sa

9、me dose to supraclavicular fossa. Bolus needed for mastectomy (and some lumpectomy)- Bolus is to give full dose to skin, rather than build-up region dose(15%-30% at entrance) to skin.- skin dose tolerance is low, and individual dependent. Use bolusevery other day, or use it daily until moist desquam

10、ation occurs.- SupraClavicular fields:* one AP field or two AP/PA field.* doctor define the field range, mainly for lymph nodes.* lymth node is usually 3cm depth à dose normalization point.* dose normalization point (ICRU 50/62/83):a) not in bone, not in air, not in fatty tissueb) 2cm away from

11、 field boarder, MLC boarder, interface of bone/tissue, interface of air/tissue, etc. (AAA à Accuros)* exit dose <70% to avoid skin reaction (entrance dose 15%-30% not a problem, exit dose could be high).* may need to rotate beam to avoid shooting spinal cord!- matching of tangential field an

12、d SupraClavicular field (to avoid cold spot and hot spot):* dose near junction area has to be not cold and not hot!* most of cases, we use half-beam-block fields for tangential fields and SupraCalvicular field(s).1.3)二. Pinnacle/Eclipse TPS Commissioning 實(shí)操2.1) Eclipse TPS data input (AAA) and beam

13、ming:- 每個(gè)能量：PDD (2x2 40x40), Profile (in-plane and cross-plane for various field sizes, variousdepths:dmax 30cm), OF (output factor) for each field size (2x2 40x40), in-air OF, etc.MLCinterleaf transmission, MLC intraleaf transmission, DLG, etc.- Eclipse Beam mling is more error forgiving, its autom

14、ated ming seems weight more on its your data is off too much fromrepresentative data, and weight less on customer data especially ifrepresentative data. It seems much less depends on commissioning physicists experience, comparedto Pinnacle beam mling.A few iterations of mling give you a good beam m.

15、- Try use EDW instead of physical wedge, which requires your commissioning that is comparable to a new beam.2.2) Pinnacle Data Input (CCC):- 每個(gè)能量: same as the data that is required for Eclipse TPS, but measure profiles with various MLC field size (not jaw field size).- playing with 30+ parameters of

16、 beam m, more than a hundred manually highly depends on the measureding erfered iterations give you a good m. Mdata accuracy and the experience of beam m- Virtually need to compare m2.3) Pinnacle CCC mwith measured data after each iteration.is expected to be not perfect (in my opinion):

17、ideally the mshould be dual source based mthat is not implemented in Pinnacle TPS;the massume the approximately 30 parameters are the same for all field sizes, while inrealitysome parameters should change with field sizes.the Pinnacle TPS assume the tray opening pointing to X-axis or inplane, while

18、for Elekta machine the trays are pointing to Y-axis or crossplane (Pinnacle match better with Varian machine in this case)the penumbra is med by the source size in X and Y directions, not considering the upper jaw andlower jaw position contributions.2.4) Pinnacle Beam Mling 實(shí)操:2.4.1) steps:- PDD and

19、 Profile measurements need to be as accurate as possible. Check data integration, e.g. PDDs for different field size are reasonably separated; profiles penumbra make sense for in-plane and cross-plane. If in doubt, discard data and re-measure.- Define all parameters related to your machine: jaws, co

20、uch, collimator, gantry, MLC, etc.- Depth Dose: adjust energy spectrum to make calculated PDD match measured PDD first, before move on to next step.- Buildup: mainly to melectron contaminations and dose at shallow depth. Can be left to the last.- Out Field: use 1cm source size in cross-plane and 0.1

21、cm source size in in-plane to start with, adjust manually to best match measured penumbra. Choose appropriate Gaussian height and width, and jaw and MLC transmission factor, from reference #1 to get best out-of-field profile (tail) match.- In Field: Arbitrary Profile parameters to m parameters to ma

22、tch measured profiles.flattening filter (better than previous cone m). Spends hours after hours to adjust- Fine tune field sizes, rounded leaf-end-gap location, etc may also help.2.4.2) My experience: a) match PDD first, then profile. B) need to re-center all your profiles. C) need tothe following f

23、irst paper tounderstand each parameters affect what part of your m. D) pay attention to some parameters I used in the screenshots, especiallythe source size, and the shape of fluence filter profile. E) according to a paper: MLC ming (leaf end and tongue & groove effect) isthe biggest impact fact

24、or to individual beam IMRT Qsing rate. F) when measuring profile, make yourself 100% sure which isin-plane profile and which is cross-plane profile. Dont mess up!2.4.3) Good References:- George Starkschall et al. Beam-commissioning methodology for a threedimensional convolution/superposition photon

25、dose algorithm. Journal of Applied Clinical Medical Physics, vol 1, No. 1, winter 2000.- Pinnacle 9.2 - Physics Instruction for Use.pdf(refer to your own Pinnacles version)- Pinnacle 9.2 - Physics Reference Guide.pdf(refer to your own Pinnacles version)- After finished the beam ming in Pinnacle TPS,

26、 you need tothe “Plan Evaluation Tools” chapter in “Pinnacle Planning using QA phantom measurements. The TG-53 report on qualityInstruction for Use 9.2” to learn how to verify the approved beam massurance of treatment planning is needed for acceptance criteria of beam m.3.1）Physics Ming (must be ran

27、dom!)- pseudo random numbers- entering photons: energy (MeV, not MV! ), location (x,y,z), orientation (, ), fluence.- random physics interaction probabilities: Rayleigh effect, P.E. effect, Compton effect, P.P. effect, etc.*. e.g. 100 KeV photon, probability of Compton effect = 0.2 / (0.2+0.4+0.001)

28、 = 33.27787%- for each type of interaction:*. Simulate probability of each possible interactions: e.g. 100+ possibilities in P.E. effect, depends on photon interaction with K shell, L shell, M shell, etc.*. Simulate resulting photon according to probability (e.g. 花生圖).- repeat above steps for every

29、and each photon or electron, primary or newly generated, until it exits material or deposited energy.三.基本原理3.2）Machine Ming- very accurate dimensions and material contents of every parts inlinac head.- phase space file can be generated before MC calculation, for all thefixed parts in Linac: target,

30、exit window, flattening filter, primarycollimator, dual chamber, etc.- MC calculation for other parts in linac head have to be done on thefly: X jaws, Y jaws, MLC, etc.- MC calculation in patient body.3.3) MC MVerification3.3.1) using some simple slabs of different material and measure dose distribu

31、tion along CAX, and compare to MC result.3.3.2) limited verifications and potential programerrors in MC code may result in dose errors for some cases.3.3.3) No complete verifications available for MC code, but it tend to be gold standard in medical physics world.4.1) Cost function:- A cost function

32、is a translation of clinical criteria.- Total cost E = Ei,tumor + Ei,tissue- Ei,tissue = (weighting factor)*(D D0), D0 is tolerance dose.Similar definition of Ei,tumor, DIY (Yourself)4.2) Simulated Annealing (SA):- So many optimization algorithms exist. SA is the most common one, and easy to underst

33、and.- SA is also widely used in many engineering and scientific fields, where its name comes from: 模擬退火- Goal: to minimize a dedicated objective function (or cost function E) by, in our case, randomly increasing or decreasingeach beamlet intensity by 0.1 cGy at every beamlet of all IMRT fields.- Aft

34、er every and each increase or decrease of beamlet intensity, the value of E is calculated, until Emin is found.四. 逆向計(jì)劃原理及目標(biāo)方程- In order not to be trapped in any “l(fā)ocal minimum”, a common problem for any optimization algorithm, we cant just ignore a unit change of beamlet intensity even if it causes

35、the E value higher. Instead, we assign a probability to this unit change, until we find: 1) another unit change that causes E to decrease, or 2) find a unit change of beamlet intensity that has minimum probability:P(E)=exp-E/T(k),P(E) is the probability caused byE.T(k) is the pseudo-temperature.- T(

36、k) is reduced at every iteration for fast cooling purposes according to the relation: T(k)=T0/kwhere k is the number of iterations, T0 the initial pseudo-temperature, and the speed parameter. The speedparameter was set to 0.6 empirically to optimize the result while keethe SA algorithm reasonably fa

37、st in radiationtherapy treatment plan optimization, according to J. Pouliot see below. The initial pseudo-temperature was set to the initial value of the cost function.* E. Lessard, J. Pouliot. Inverse planning anatomy-based dose optimization for HDR-brachytherapy of the prostate using fast simulate

38、d annealing algorithm and dedicated objective function. Med. Phys., 28(5)., 773-779. (2001)- The allowance for both the “downhill”and “uphill” moves of the beamlet intensities saves the algorithm from beco at local minima. It statistically guarantees the finding of an optimal solution: global minimu

39、m of cost function.stuckIllustration: how to search the minimum height on a mountain top, without being trapped at “l(fā)ocal minimum”.(Passing rate)TG 119: IMRT commissioning: Multiple institution planning and dosimetry comparisons1) H&N phantom, prostate phantom sent to 7 institutions. Phantoms we

40、re CTed, planned, delivered, and dose measured, then gamma analyzed the passing rate.: < 93%, need analysis.(optimal plan but sub-optimal delivery, or sub-optimal plan but optimal delivery)5.1）IMRT Commissioning and Verifications五. 其他前期問題答疑：< 90%, need to take action: re-plan may be needed, ML

41、C maintenance may be needed, etc.:5.2.1) For most of cases there are no significant difference.5.2.2) VMAT has faster delivery speed: 1-2 min vs 5-10 min. VMATbecomes more popular.5.2.3)IMRTmaybemoreflexibleinsomebraincases:non-coplanar field from top of head: sup/lt, sup/rt, sup/ant, sup/post,sup/l

42、t/ant, etc.5.2.4) Time spent on planning are about the same for both. VMATneed more commissioning job and QA jobs than IMRT, but shortertreatment time.5.2) More on IMRT vs. VMAT5.3.1) 各種認(rèn)證機(jī)構(gòu)：- CAMPEP (master and residency program), DABR exams/diploma, MOC (maintenance of), ACR accreditation forcancer center (American College of Radiology), JC (Joint Commission), Magnet (Nurse), etc.- IOMP, especially IMPCB: International Me