Imaging Physics for Radiation Oncology Residents- AAPM Chapter

1、New England Chapter of the AAPM2021 Winter MeetingNewton, MAFebruary 1, 2021 Making High Tech Radiotherapy SafeHoward Amols, Ph.D. Memorial Sloan Kettering Cancer Center New York, USA Radiation Therapy Accidents WHO World Alliance for Patient Safety Literature review 1976-2007 3125 patients affected

2、 by errors in radiotherapy 1% resulted in death (middle and high income countries*) 98% of errors occurred in planning stage or during introduction of new systems or equipment Of errors without adverse effects 55% planning 25% new systems or equipment 9% information transfer 10% treatment delivery M

3、is-information or errors in data transfer- greatest bulk of “near misses in modern radiotherapy servicesRadiation Therapy safety problems3 Myths nRadiation therapy accidents are rarenMost accidents happened long ago in the developing world nYou need new high technology to have an accidentTheres been

4、 a focus on high-technology as the source of RT accidents.This is only partially true. New technology is often a contributing factor, but most RT disasters involving new technology required significant help from humans in order to escalate.Really big errors are a team effort!(more than one person ha

5、s to make a mistake) A Major Accident Requires:Equipment malfunction AND/OR an individual act of negligence AND a deficient QA/QC program, bad communications, inadequate training, lack of teamwork If commercial airlines had the same error rates as hospitals there would be a major airline crash EVERY

6、 DAY.There is something fundamentally wrong with the historical method by which the medical profession has approached QA/QC:“Those who dont know history are destined to repeat it. Edmund Burke Radiation Accidents: Common Threadsnew equipment + new software, but no new QA/QC understaffed, overworked,

7、 undertrained, rushedno internal redundancy, no external auditsno common sense, no time outsbad communication, no central reportingtoo much faith in manufacturer design manufacturer and institutional denialunusual clinical results ignored by physiciansA recent IAEA analysis of reported treatment err

8、ors noted common threads: Possibility of systematic software error never considered;2. Manufacturer failed to inform users of previous incidents;3. Error messages from system ignored or overridden;5. Used software in unanticipated sequence that confused the system;6. Input wrong data, misunderstood

9、input requirements;7. No physical dose measurements made to confirm;8. New or recently upgraded equipment;Life is more complicated than it used to be! Control software for Varian Truebeam Linacr (for example) has 1 million lines of computer code! No single person can completely understand how such a

10、 system works No government regulatory agency has the expertise to review such a system Teamwork and communication is essentialRT is more complicated than it used to be! Manufacturers software engineers rarely have clinical experience and do not know how a clinic really operates People will always f

11、ind new ways to make mistakes that are difficult to anticipated Quality Assurance must be an evolving process, to be reviewed and modified whenever new technology is implemented and when errors are discovered Study near-misses as well as actual errors Never underestimate the ingenuity of a fool, or

12、someone who is being rushed!Special Dangers of Hi-Technology Systematic errors harder to detect Humans get complacent. Dont really check computers (evolution of R/V systems) Many treatment components too complex for humans to check (e.g., DMLC files, IMRT files)4. Many treatment aids/devices are inv

13、isible (dynamic wedge, MLC, etc.)5. Errors made on day 1 can propagate 6. Programmers dont understand what we do7. We dont understanding what programmers do 7. Too easy to over ride warning messages8. Manufacturers training programs often inadequateMy favorite IMRT accident Almost everything that ca

14、n go wrong in a radiation therapy accident occurred here! Best teaching tool I know of on the subject of radiation therapy errorsBackground, IMRT accidentMarch 2005, New York City A patient is to be treated with IMRT for head and neck cancer (oropharynx)March 4 7: An IMRT plan is prepared: “1Orophar

15、yn. Verification plan created by TPS. EPID dosimetry confirms correctness.March 8: patient treated correctly with “1Oropharyn. March 9-11: Fractions #2, 3 and 4 also correct. Verification images for the kV imaging system are created and added to the plan, now called “1AOropharyn.March 11: Physician

16、wants modified dose distribution (reducing dose to teeth) “1AOropharyn is copied and saved to the DB as “1BOropharynNOTE the dates!March 14: Re-optimization for “1B Oropharyn.New optimal fluences saved to DB.MLC motion control points for IMRT generated. Normal completion.What happened?March 14“Save

17、all is started. All new and modified data should be saved to the DB.In this process, data is sent to a holding area on the server, and not saved permanently until ALL data elements have been received.Data to be saved included: (1) fluence data, (2) DRRs and (3) MLC control pointsWhat happened?March

18、14, 11 a.m.An error message is displayed.The user presses “Yes, which begins a second, separate, save transaction.MLC control point data is moved to the holding area.The purpose of this error message is so that you can click yes to proceed What happened?March 14, 11.a.m.The DRR is, however, still lo

19、cked into the faulty first attempt to save. This means the second save wont be able to complete.The software would have appeared to be frozen.Ctrl-Alt-Del usually gets you out of this so.What happened?March 14, 11.a.m.Within 12 s, another workstation is used to open the patients plan to load into VA

20、RIS and to treat. Unbeknownst to the users, the MLC motion file is NOT properly saved because the error message was over-riddenWhat happened?March 14, 11 a.m.No verification plan, no pre-treatment dosimetry, no review by 2nd physicist (not enough time!)Several computer crashes ignored and over-ridde

21、n.Plan approved by physician (the plan looked OK on printout, but the computer data files were corrupted)Big Errors are usually a team effort!So far: The radiation oncologist did not read port films on time, and then rushed everybody else Neither therapists nor physicists confront him The treatment

22、planner over-rode an error message they did not understand The physicists did neither a double check, nor a dosimetry test of an IMRT plan (also called billing fraud) There were real bugs in the software design (it crashed, two people could open file at the same time) So far no real harm has been do

23、ne! But theyre sowing the seeds for a really great accidents! And now the therapists add the final piece!March 14, 2005, 1 p.m.What they should have seen:March 14, 2005, 1 p.m.: What they also didnt notice:1324 MU with MLC wide openDiscovery of accidentMarch 14-16, 2005The patient is treated without

24、 MLCs for 3 fractionsOn March 16, a verification plan is created and run on the treatment machine. The operator notices the absence of MLCs.A second verification plan is created and run with the same result.The patient plan is loaded and run, with the same result.Impact of accidentThe patient receiv

25、ed 13 Gy per fraction for three fractions, i.e. 39 Gy in 3 fractionsWas there a bug in the Manufacuters software?Yes, but the software bug needed a lot of help in order to kill somebody: MD didnt check port films and rushed the plan modification Nobody confronted the MD Error messages ignored and no

26、t investigated Treatment plan QA not performed Therapists werent watching MLC display neither the manufacturer, nor the user, nor the FDA, nor anyone else has the expertise to test this software for safety6. Accident not made public until much later! Disasters are a team effort! More radiation thera

27、py accidentsSpringfield, Mo., 2004-09. 76 SRS patients overdosed 50 percent. Used too large a dosimeter to calibrate SRS fields. No independent check, no mandated state or federal reporting reguirement, no requirement for physicists or therapists to be certified. Insufficient support from manufactur

28、er.Tampa, Fla., 2004-5. , 77 SRS patients overdosed 50% because PDD factor not used in TG-51 calibration. Uncovered after 1 year during RPC inspection for participation in RTOG.10mm24mm7mmFor 10mm coneThe Farmer chamberis much larger than the x-ray field, and most of it is in the beam penumbra.Dose

29、reading = (6+4x0.7)/24 = 37%Penumbra= 3-5 mmPenumbra= 3-5 mmNote: dose uniformity should be +3% over 80% of the field width (for 10 x 10 field size).80% of field = 8mm. But for very small fields uniformity is worse.More radiation therapy accidentsAnother radiosurgery error: Linac collimators not pro

30、perly set for small SRS cones (hospital staff assumed this was done automatically)SRS coneInsert (2-3)SRS coneHolder(5x5)Linac Collimators(10 x10)More.UK, 1982-90: incorrect SSD correction (did not know how TPS worked). 1045 patients, 30% underdose, 492 RT failuresBend, Oregon, 1980s: incorrect TPC.

31、 13% overdoseSpain, 1990: Linac repair led to 36MeV e- beam no matter what was programmed. No dosimetry check. 27 patients, 15 deathsFrance, 2004: incorrect MU for dynamic wedge. 23 patients overdosed 20%, 4 deathsGlasgow,2006: incorrect calculation of MU. Planner thought TPS calculated MU/Gy and no

32、t MU/fraction. It didnt! 67% overdose results in deathFrance, 2006-7: large ion chamber used for SRS. 145 overdoses.Errors we have Seen at Memorial Sloan KetteringMLC Error (FSRT) Event Sequence: FSRS treatment scheduled right after DMLC treatment At end of DMLC treatment leaves are all closed Brain

33、Lab mMLC attached to Linac below regular MLC Therapist does not retract primary MLC leaves Primary MLC leaf position not detected by R&V (thinks its SRS) Patient treated with correct mMLC apertures and closed primary MLCWhy: Software not designed for two MLCs Light field not normally used for SRSDML

34、C Error: IMRT treatment with open MLC leaves:DMLC field selected for treatment after a static MLC treatmentDMLC plan loaded, leaves retracted for light field use“Go selected, leaves fail to return to prescribed positionAll systems allow treatment to proceed with retracted leavesTherapist fails to de

35、tect error Why:Software did anticipate this sequence of eventsCould only have happened with exactly the right wrong timing Very similar to Therac-25 disaster Record/Verify Systematic Error with DMLC: DMLC (v1) created from TPS and sent to R/V Dosimetry checks done Plan changed MU (only) manually edi

36、ted in R&V No Independent Check of Data Difference too subtle to see on Portal Image Check Why: Human error Improper understanding of software (change in TP system doesnt automatically get transferred to R/V)Upgraded,VARISto ARIARadiation Therapy Safety Program at MSKCC MSK was the first center in t

37、he US to initiate IMRT Since 1995 we have treated over 20,000 patients with IMRT ( half a million treatment fractions) We never dismiss the potential for errors, and have put into place a multi-tiered Quality Assurance Program to continually monitor all phases of radiation therapy treatments at MSK

38、Almost all radiation therapy accidents result from a combination of equipment (or software) malfunctions PLUS human error Radiation Therapy Safety Program at MSKCC- training in all new technologies- staff all board certified and licensed- continuing staff education - all errors, large and small are

39、fully discussed at monthly meetings of the Rad Onc QA Committee- Minutes of Rad Onc QA Committee are reported and discussed at the Hospitals QA Committee- QA program is continually modified and upgraded to reflect new findings and technologies- New equipment and computer systems are tested prior to

40、release for patient treatmentsThe human component of most radiation therapy accidents results from a combination of: failure to appreciate limitations of new technologies inadequate redundency in QA program:- people checking people- people checking computers- computers checking people- computers che

41、cking computers static or non-proactive QA program System wide QA of all treatment devices, hardware and software Patient specific QA for each individual patients customized treatment plan Quality Assurance at MSK consists of two equally important components :System wide QA of all treatment devices,

42、 hardware and software Annual dosimetry intercalibration test with Radiological Physics Center Periodic dosimetry and treatment planning accreditation from RTOG for IMRT and IGRT national protocol studies Periodic reviews by ACR (ASTRO?) Every day, during each treatment fraction for all patients all

43、 treatment parameters (dose, mlc settings, x-ray energy, etc. are recorded by the linac computer and stored in a treatment log file Every evening a batch computer program is run comparing all IMRT treatment log files with the planned MLC files. Discrepancies are investigated immediately by a medical

44、 physicist On a weekly basis all patient treatment folders are reviewed by both a medical physicist and a radiation therapist to insure consistency between delivered treatments and treatment plans On a weekly basis portal films and/or orthogonal x-rays are taken on each patient to insure correctness

45、 of patient treatment position Using Linac Log FilesHuman errors usually fall into one of three categories:1. Department policy is properly followed, but an error occurs anyway. For example, policy requires that all treatment plans and MU calculations be independently checked by a second person befo

46、re the patients first treatment. This policy is followed, but the second person also fails to detect the error;2. Department policy is not followed. For example, no one performs an independent check of treatment plan or dose calculation prior to the first treatment; and3. Department policy is defici

47、ent or incomplete. Most common for new technology4. Bizarre errors: sequence of events, almost impossible to foresee or preventNothing is foolproof for the sufficiently talented fool!R/V systems, computer controlled Linacs, image guided patient positioning systems, etc. reduce but do no preventerror

48、s. They enable humans to make different kinds of mistakes faster and more efficiently.New Paradigm for QAMost errors are NOT systematic. They are patient specific. Therefore QA should shift from equipment focused to patient focused. Patient Specific QA: Treatment plan check (more difficult than befo

49、re)R/V, file check-sums (each fraction)Independent MU check, dosimetry, portal imagesLog file analysis, chamber measurement, film dosimetry Machine Specific QA:Film test Dosimetry test Drift test MLC and IGRT tests Is QA Reporting Like an Iceberg?ReportedErrors Discovered Errors Actual number of err

50、orsThe Aviation ApproachEven good people makemistakes, they are the the rule, not the exceptionMistakes result from flaws in the systemFind the system flaw that made it possible for a particular individualto make a mistake Start at the top and workdown the system until thecause of the error is found

51、 They rarely make the sameMistake twiceThe Hospital ApproachWe are Gods. Mistakes are the exceptions not the rule Mistakes result from individual screw upsFind and punish the person who made the mistake Start at the bottom and work up. Find the lowest ranking person you can blame The Aviation Approa

52、chEven good people makemistakes, they are the the rule, not the exceptionMistakes result from flaws in the systemFind the system flaw that made it possible for a particular individualto make a mistake Start at the top and workdown the system until thecause of the error is found They rarely make the

53、sameMistake twiceThe Hospital ApproachWe are Gods. Mistakes are the exceptions not the rule Mistakes result from individual screw upsFind and punish the person who made the mistake Start at the bottom and work up. Find the lowest ranking person you can blame The Aviation ApproachEven good people mak

54、emistakes, they are the the rule, not the exceptionMistakes result from flaws in the systemFind the system flaw that made it possible for a particular individualto make a mistake Start at the top and workdown the system until thecause of the error is found They rarely make the sameMistake twiceThe H

55、ospital ApproachWe are Gods. Mistakes are the exceptions not the rule Mistakes result from individual screw upsFind and punish the person who made the mistake Start at the bottom and work up. Find the lowest ranking person you can blame The Aviation ApproachEven good people makemistakes, they are th

56、e the rule, not the exceptionMistakes result from flaws in the systemFind the system flaw that made it possible for a particular individualto make a mistake Start at the top and workdown the system until thecause of the error is found They rarely make the sameMistake twiceThe Hospital ApproachWe are

57、 Gods. Mistakes are the exceptions not the rule Mistakes result from individual screw upsFind and punish the person who made the mistake Start at the bottom and work up. Find the lowest ranking person you can blame The Aviation ApproachEven good people makemistakes, they are the the rule, not the ex

58、ceptionMistakes result from flaws in the systemFind the system flaw that made it possible for a particular individualto make a mistake Start at the top and workdown the system until thecause of the error is found They rarely make the sameMistake twiceThe Hospital ApproachWe are Gods. Mistakes are th

59、e exceptions not the rule Mistakes result from individual screw upsFind and punish the person who made the mistake Start at the bottom and work up. Find the lowest ranking person you can blame The Aviation ApproachEven good people makemistakes, they are the the rule, not the exceptionMistakes result from flaws in the systemFind the system flaw that made it possible for a particular individualto make a mistake Start at the top and workdown the system until thecause of the error is found They rarely make the sameMistake twiceThe Hospital ApproachWe are Gods. Mista