As clinical medical physicists, we are applied scientists helping our fellow clinical practitioners (physicians who have completed professional training in the practice of medicine) to safely and effectively practice in “science‐heavy” subspecialties of medicine. We are uniquely qualified to bring science into the clinical practice in the appropriate context. To do our jobs effectively, we must “know the trade” of our physician colleagues (hence the term clinical medical physicist), but we must not lose sight of our role as the sole scientist in the endeavor. Toward that end, I would like to encourage us to clearly acknowledge uncertainties. Doing so is the right thing to do scientifically, and also serves the patient's interests — our primary ethical obligation. I firmly believe that when we clearly acknowledge the uncertainties in a process, we not only serve the patient's interests appropriately but we also enhance the physician's awareness of inherent limitations and, if done appropriately, enhance the stature of the medical physics profession.

Allow me to provide some context based on my personal experience as a clinical radiotherapy physicist. I believe the overall theme would apply similarly in the diagnostic imaging and nuclear medicine specialties.

We have access to exquisite image data, sophisticated image registration algorithms, automated tissue segmentation models, and powerful dose calculation algorithms. Consequently, we risk falling prey to “false accuracy”.1 In its most benign form, this results in reduced efficiency as the planner expends time and effort to achieve a very small shift in a particular dosimetric parameter in order to be “under tolerance.” In a less benign form, this can result in suboptimal target coverage or suboptimal normal‐tissue dose.

In my own institution, I recently completed a review of hundreds of patient charts as part of a comprehensive re‐evaluation of our treatment planning processes, and discovered significant variation in how "organs at risk" (OARs) are contoured. Some of these have a substantive impact on how plans are optimized given the particular dose–volume objectives used. Interobserver variation in contouring has been well demonstrated in the literature, even for “well defined” organs.2-5 When the dosimetric objective is mean dose or a relative dose–volume metric, the variation in contouring can introduce significant uncertainty. As described in Yock et al,1 the dosimetric impact has been demonstrated to be as much as 5% for clinically relevant uncertainties. In our institution's contour review, we observed a mean heart volume of 600 cc with a standard deviation of 300 cc due to differences among planners in the extent of superior pericardium contoured — yet the main dosimetric objective for non‐SBRT plans is mean dose. [This was addressed through standardized contouring guidelines.]

Speaking of the dose objectives, many "tolerance doses" used for both conventionally fractionated and hypofractionated treatments are not based on solid clinical data but are largely the preferences or practices of prominent authors.

Quoting from the QUANTEC Science Overview6: “Dose–volume constraints are used in routine dose planning as an integral part of the informal optimization of therapeutic ratio that inverse planning entails. Acceptable dose distributions are identified from an assessment of the risk:benefit ratio in an individual patient—often on the basis of clinical experience rather than on numerical estimates from dose–volume models. Population constraints are very important in this context but can obviously not stand alone. Careful consideration should be given not only to the numerical value of these constraints but also to their statistical uncertainty. Using these values directly in dose–plan optimization should be done with great caution.

There is still a lack of proper estimation of the uncertainty in these parameters in most cases.”

From the TG‐101 section on normal tissue dose tolerance7: The doses are mostly unvalidated, and while most are based on toxicity observation and theory, there is a measure of educated guessing involved as well.

Modern dose calculation algorithms are quite impressive, but rely on CT Hounsfield numbers to infer the material composition of the medium. Other examples of common uncertainties include, but are not limited to, deformable and rigid image registration,8 applicability of OAR dose objectives when combining different fractionation regimens and/or previous treatment,9 motion management uncertainties, peripheral target coverage with single‐isocenter multitarget techniques, or 4D binning artifacts from irregular breathing patterns. The list could go on and on.

My point is that if we as clinical physicists do not explore such sources of uncertainty and clearly explain them to our physician colleagues, we are doing our physician colleagues (and by extension their patients) a disservice by not understanding how these factors interrelate to impact the patient's care. We should help our clinical colleagues to appreciate the uncertainties in complex processes so they can better integrate the uncertainty into the management of their patients' needs.

作为临床医学物理学家，我们是应用科学家，旨在帮助我们的临床同胞（已完成医学实践专业培训的医学家）安全有效地从事“重科学”医学专业的实践。我们具有在适当情况下将科学纳入临床实践的独特资格。为了有效地完成我们的工作，我们必须“了解”我们的医师同事的职业（因此称为“临床医学物理学家”），但是我们绝不能忽略我们作为唯一科学家的角色。为此，我谨鼓励我们明确承认不确定因素。这样做是科学上正确的做法，也符合患者的利益-我们的主要道德义务。我坚信，当我们清楚地认识到过程中的不确定性时，我们不仅会适当地服务于患者的利益，而且还会增强医师对内在局限性的认识，并且，如果采取适当的措施，也会增强医学物理学专业的地位。

请允许我根据我作为临床放射治疗物理学家的经验提供一些背景信息。我相信整个主题将同样适用于诊断成像和核医学专业。

我们可以访问精美的图像数据，复杂的图像配准算法，自动组织分割模型以及强大的剂量计算算法。因此，我们冒着沦为“错误准确性”的猎物的风险。1以最良性的形式，这会导致效率降低，因为计划人员会花费大量时间和精力来实现特定剂量参数的很小变化，以达到“承受能力差”的目的。以较温和的形式，这可能会导致靶标覆盖率不足或正常组织剂量不足。

在我自己的机构中，我最近完成了对数百张患者图表的审查，这是对我们的治疗计划流程进行全面重新评估的一部分，并且发现“危险器官”（OAR）的轮廓存在显着差异。考虑到所使用的特定剂量-体积目标，其中一些会对计划的优化方式产生实质性影响。观察者之间轮廓的变化在文献中已得到充分证明，即使对于“定义明确的”器官也是如此。2-5当剂量学目标是平均剂量或相对剂量-体积度量标准时，轮廓变化会带来很大的不确定性。如Yock等人所述1对于临床相关的不确定性，剂量影响已被证明高达5％。在我们机构的轮廓审查中，由于上级心包轮廓的规划者之间存在差异，我们观察到平均心脏容量为600 cc，标准差为300 cc，但非SBRT计划的主要剂量目标是平均剂量。[这已通过标准化轮廓指南解决。]

说到剂量目标，用于常规分次治疗和次分次治疗的许多“耐受剂量”并不是基于可靠的临床数据，而是很大程度上是著名作者的偏爱或实践。

引用QUANTEC科学概述6：“在常规剂量计划中使用剂量-体积限制，这是逆向计划所需要的非正式治疗比例优化的组成部分。可以通过评估单个患者的风险：获益比率来确定可接受的剂量分布，通常是基于临床经验，而不是基于剂量-体积模型的数字估计。在这种情况下，人口限制非常重要，但显然不能独立存在。不仅应仔细考虑这些约束的数值，而且还应考虑其统计不确定性。在剂量计划优化中直接使用这些值时应格外小心。

在大多数情况下，仍缺乏对这些参数不确定性的适当估计。”

从TG-101部分的正常组织剂量耐受性7开始：剂量大多未经验证，尽管大多数是基于毒性观察和理论，但也有一定程度的有根据的猜测。

现代的剂量计算算法令人印象深刻，但是依靠CT霍恩斯菲尔德数来推断介质的材料组成。常见不确定性的其他示例包括但不限于可变形和刚性图像配准，结合不同分级方案和/或先前治疗时8的OAR剂量目标的适用性，9运动管理不确定性，单等中心多目标技术的外周目标覆盖率或来自不规则呼吸模式的4D分箱伪像。该清单可以继续下去。

我的观点是，如果我们作为临床物理学家不探究此类不确定性来源，并向我们的医师同事清楚地解释它们，那么我们正在通过不了解这些因素如何相互影响来影响医师同事（并扩展他们的患者），从而造成损害。病人的护理。我们应该帮助临床同事了解复杂过程中的不确定性，以便他们可以更好地将不确定性纳入患者需求的管理中。