Search

ImportanceThere is a growing interest in understanding whether cardiovascular magnetic resonance (CMR) myocardial tissue characterization helps identify risk of cancer therapy–related cardiac dysfunction (CTRCD).ObjectiveTo describe changes in CMR tissue biomarkers during breast cancer therapy and their association with CTRCD.Design, Setting, and ParticipantsThis was a prospective, multicenter, cohort study of women with ERBB2 (formerly HER2)–positive breast cancer (stages I-III) who were scheduled to receive anthracycline and trastuzumab therapy with/without adjuvant radiotherapy and surgery. From November 7, 2013, to January 16, 2019, participants were recruited from 3 University of Toronto–affiliated hospitals. Data were analyzed from July 2021 to June 2022.ExposuresSequential therapy with anthracyclines, trastuzumab, and radiation.Main Outcomes and MeasuresCMR, high-sensitivity cardiac troponin I (hs-cTnI), and B-type natriuretic peptide (BNP) measurements were performed before anthracycline treatment, after anthracycline and before trastuzumab treatment, and at 3-month intervals during trastuzumab therapy. CMR included left ventricular (LV) volumes, LV ejection fraction (EF), myocardial strain, early gadolinium enhancement imaging to assess hyperemia (inflammation marker), native/postcontrast T1 mapping (with extracellular volume fraction [ECV]) to assess edema and/or fibrosis, T2 mapping to assess edema, and late gadolinium enhancement (LGE) to assess replacement fibrosis. CTRCD was defined using the Cardiac Review and Evaluation Committee criteria. Fixed-effects models or generalized estimating equations were used in analyses.ResultsOf 136 women (mean [SD] age, 51.1 [9.2] years) recruited from 2013 to 2019, 37 (27%) developed CTRCD. Compared with baseline, tissue biomarkers of myocardial hyperemia and edema peaked after anthracycline therapy or 3 months after trastuzumab initiation as demonstrated by an increase in mean (SD) relative myocardial enhancement (baseline, 46.3% [16.8%] to peak, 56.2% [18.6%]), native T1 (1012 [26] milliseconds to 1035 [28] milliseconds), T2 (51.4 [2.2] milliseconds to 52.6 [2.2] milliseconds), and ECV (25.2% [2.4%] to 26.8% [2.7%]), with P Conclusions and RelevanceIn women with ERBB2-positive breast cancer receiving sequential anthracycline and trastuzumab therapy, CMR tissue biomarkers suggest inflammation and edema peaking early during therapy and were associated with ventricular remodeling and BNP elevation. However, the increases in CMR biomarkers were transient, were not associated with LVEF or myocardial strain, and were not useful in identifying traditional CTRCD risk.

Machine learning (ML) holds great promise for impacting healthcare delivery; however, to date most methods are tested in ‘simulated’ environments that cannot recapitulate factors influencing real-world clinical practice. We prospectively deployed and evaluated a random forest algorithm for therapeutic curative-intent radiation therapy (RT) treatment planning for prostate cancer in a blinded, head-to-head study with full integration into the clinical workflow. ML- and human-generated RT treatment plans were directly compared in a retrospective simulation with retesting (n = 50) and a prospective clinical deployment (n = 50) phase. Consistently throughout the study phases, treating physicians assessed ML- and human-generated RT treatment plans in a blinded manner following a priori defined standardized criteria and peer review processes, with the selected RT plan in the prospective phase delivered for patient treatment. Overall, 89% of ML-generated RT plans were considered clinically acceptable and 72% were selected over human-generated RT plans in head-to-head comparisons. RT planning using ML reduced the median time required for the entire RT planning process by 60.1% (118 to 47 h). While ML RT plan acceptability remained stable between the simulation and deployment phases (92 versus 86%), the number of ML RT plans selected for treatment was significantly reduced (83 versus 61%, respectively). These findings highlight that retrospective or simulated evaluation of ML methods, even under expert blinded review, may not be representative of algorithm acceptance in a real-world clinical setting when patient care is at stake. An artificial intelligence system prospectively deployed to design radiation therapy plans for patients with prostate cancer illustrates the real-world impact of machine learning in clinical practice and identifies factors influencing human–algorithm interaction

Spatially resolved ambient mass spectrometry imaging methods have gained popularity to characterize cancer sites and their borders using molecular changes in the lipidome. This utility, however, is predicated on metabolic homogeneity at the border, which would create a sharp molecular transition at the morphometric borders. We subjected murine models of human medulloblastoma brain cancer to mass spectrometry imaging, a technique that provides a direct readout of tissue molecular content in a spatially resolved manner. We discovered a distance-dependent gradient of cancer-like lipid molecule profiles in the brain tissue within 1.2 mm of the cancer border, suggesting that a cancer-like state progresses beyond the histologic border, into the healthy tissue. The results were further corroborated using orthogonal liquid chromatography and mass spectrometry (LC-MS) analysis of selected tissue regions subjected to laser capture microdissection. LC-MS/MS analysis for robust identification of the affected molecules implied changes in a number of different lipid classes, some of which are metabolized from the essential docosahexaenoic fatty acid (DHA) present in the interstitial fluid. Metabolic molecular borders are thus not as sharp as morphometric borders, and mass spectrometry imaging can reveal molecular nuances not observed with microscopy. Caution must be exercised in interpreting multimodal imaging results stipulated on a coincidental relationship between metabolic and morphometric borders of cancer, at least within animal models used in preclinical research.

We have applied a new simplified combination of numerical methods for studying the time and three-dimensional space dependence of quasi-three-level Yb3+:Yttrium Aluminum Garnet (YAG) end-pumped lasers passively Q-switched by a Cr4+:YAG saturable absorber. We base our 3-D model on iterative, efficient, time- and space-dependent numerical propagation of the optical field through the laser cavity. The complex-valued laser field is coupled to the Yb3+:YAG and Cr4+:YAG media via complex optical permittivities, which are subsequently altered by gain/loss intensity saturation. The calculation is simplified using the radial symmetry of the system, with the cavity round-trip time as the smallest increment for updating the permittivities. We also include the effects of field diffraction in an intra-cavity air gap. For specified CW spatial pump conditions, self-consistent repetitively pulsed solutions for the laser field in a flat-flat or flat-convex mirror cavity are found with no ad hoc laser mode size or shape assumptions; these solutions are not Gaussian modes. We concentrate on compact lasers with multi-Watt average output power, operating at modest pulse energy (~1.0 mJ), high repetition rate (~5 kHz) and short pulse duration (~1.5 ns). Typical room-temperature pump-to-laser slope power efficiencies exceeding 50% are predicted, depending on laser pump and cavity loss parameters. Model results agree well with recently published experimental data.

Purpose: Partial Breast Irradiation (PBI) treats the tumor bed plus a margin in low risk early stage breast cancers, reducing exposure to organs at risk. Randomized trials have shown that PBI with external beam radiotherapy is non-inferior to whole breast irradiation in terms of local control with fewer late normal tissue adverse events. We aimed to determine the proportion of women with early stage breast cancer suitable for external beam PBI at our institution using adapted ASTRO guidelines for accelerated PBI whereby both clinical-pathological factors and visualization of the tumor bed/cavity on planning scans were determined to quantify PBI eligibility. Methods: Patients were identified retrospectively from January 1, 2013 to November 1, 2018 based on age (≥55), tumor size (≤2cm invasive, ≤2.5cm DCIS), nodal negativity and grade (1 or 2). Cases identified were unifocal, lymphovascular invasion negative, ER positive, HER2 negative, with margins ≥2mm (invasive) and ≥3mm (DCIS), no chemotherapy, non-lobular, no extensive intraductal component, no previous/ bilateral cancers and endocrine therapy optional. Planning scans were reviewed to assess for presence of surgical clips and assigned cavity visualization scores (CVS); CVS 1=cavity not visualized, CVS 2=cavity visualized with indistinct margins, CVS 3=cavity visualised with some distinct margins and heterogeneous appearance, CVS 4= cavity visualised with mostly distinct margins and mildly heterogeneous appearance, CVS 5= all cavity margins clearly defined and homogeneous appearance. These features were used to determine if cases were considered radiologically suitable for PBI. Results: Thirty-seven percent (1051/2824) of cases identified were deemed eligible based on age, grade and T1N0/Tis. Of the 1051, 306 (29%) met all the pathologic criteria; the main reasons for exclusion were previous breast cancer, lobular histology, or multifocality. Median age of eligible cases was 65 years (range 55-88 years). The majority (254) had invasive disease (T1mi: 6, T1a: 17, T1b: 98, T1c: 133); 52 had DCIS. There were 169 right-sided and 137 left-sided cases. Endocrine therapy was documented in 55% of cases. CVS (1-5) were: 11%, 22%, 30%, 29%, 8%. Overall, 66% (201/306) of cases were considered radiologically suitable for PBI (of the 201 cases, 97% had CVS ≥3). Seventy-two cases (72/306; 24%) had tumor bed clips, of which, 34 were considered suitable for PBI (91% CVS ≥3; 9% CVS 2). Conclusions: Overall, a low proportion of all patients were eligible for PBI based on clinical features, but two-thirds of eligible cases were suitable for PBI based on CVS score alone. In contrast, the suitability for PBI within the cohort with surgical clip placement was low. Visualization of the tumor bed is essential to safely deliver PBI and is guided by identification of the cavity seroma and/or surgical clips. Development of standard practice guidelines for surgical clip positioning to identify the tumor bed will potentially assist in increasing the proportion of early stage breast cancer patients suitable to receive PBI, even if the CVS is less favorable. Citation Format: Christine Anne Koch, Gemma Corey, Kathy Han, Grace Lee, Tom Purdie, Chris McIntosh, Patricia Lindsay, Fei-Fei Liu, Anthony Fyles, Wilf Levin, Jennifer Croke, Aisling Barry, Danielle Rodin, Nhu Tram Nguyen, Tatiana Conrad. Evaluation of partial breast irradiation suitability in early stage breast cancer patients [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P4-12-24.

BACKGROUND: Coronary heart disease (CHD), heart failure (HF) and stroke are complex diseases with multiple phenotypes. While many risk factors for these diseases are well known, investigation of as-yet unidentified risk factors may improve risk assessment and patient adherence to prevention guidelines. We investigated the diet domain in the Framingham Heart Study (FHS), Cardiovascular Heart Study (CHS), and Atherosclerosis Risk In Communities (ARIC) study to identify potential lifestyle and behavioral factors associated with CHD, HF, and stroke. METHODS: We used machine learning feature selection based on random forest analysis to identify potential risk factors associated with CHD, stroke, and HF in FHS. We evaluated the significance of selected variables using univariable and multivariable Cox proportional hazards analysis adjusted for known CV risks. Findings from FHS were then validated using CHS and ARIC. RESULTS: We identified multiple dietary and behavioral risk factors for CVD outcomes including marital status, red meat consumption, whole milk consumption, and coffee consumption. Amongst these dietary variables, increasing coffee consumption was associated with decreasing long-term risk of HF congruently in FHS, ARIC, and CHS. CONCLUSIONS: Higher coffee intake was found to be associated with reduced risk of HF in all three studies. Further study is warranted to better define the role, possible causality, and potential mechanism of coffee consumption as a potential modifiable risk factor for HF.

Radiotherapy is a common treatment modality for the treatment of cancer, where treatments must be carefully designed to deliver appropriate dose to targets while avoiding healthy organs. The comprehensive multi-disciplinary quality assurance (QA) process in radiotherapy is designed to ensure safe and effective treatment plans are delivered to patients. However, the plan QA process is expensive, often time-intensive, and requires review of large quantities of complex data, potentially leading to human error in QA assessment. We therefore develop an automated machine learning algorithm to identify ‘acceptable’ plans (plans that are similar to historically approved plans) and ‘unacceptable’ plans (plans that are dissimilar to historically approved plans). This algorithm is a supervised extension of projective adaptive resonance theory, called SuPART, that learns a set of distinctive features, and considers deviations from them indications of unacceptable plans. We test SuPART on breast and prostate radiotherapy datasets from our institution, and find that SuPART outperforms common classification algorithms in several measures of accuracy. When no falsely approved plans are allowed, SuPART can correctly auto-approve 34% of the acceptable breast and 32% of the acceptable prostate plans, and can also correctly reject 53% of the unacceptable breast and 56% of the unacceptable prostate plans. Thus, usage of SuPART to aid in QA could potentially yield significant time savings.

The complexity of generating radiotherapy treatments demands a rigorous quality assurance (QA) process to ensure patient safety and to avoid clinically significant errors. Machine learning classifiers have been explored to augment the scope and efficiency of the traditional radiotherapy treatment planning QA process. However, one important gap in relying on classifiers for QA of radiotherapy treatment plans is the lack of understanding behind a specific classifier prediction. We develop explanation methods to understand the decisions of two automated QA classifiers: (1) a region of interest (ROI) segmentation/labeling classifier, and (2) a treatment plan acceptance classifier. For each classifier, a local interpretable model-agnostic explanation (LIME) framework and a novel adaption of team-based Shapley values framework are constructed. We test these methods in datasets for two radiotherapy treatment sites (prostate and breast), and demonstrate the importance of evaluating QA classifiers using interpretable machine learning approaches. We additionally develop a notion of explanation consistency to assess classifier performance. Our explanation method allows for easy visualization and human expert assessment of classifier decisions in radiotherapy QA. Notably, we find that our team-based Shapley approach is more consistent than LIME. The ability to explain and validate automated decision-making is critical in medical treatments. This analysis allows us to conclude that both QA classifiers are moderately trustworthy and can be used to confirm expert decisions, though the current QA classifiers should not be viewed as a replacement for the human QA process.

Recent technologies have enabled us to collect diverse types of genome-wide data at an unprecedented scale and in multiple dimensions. Integrative computational methods are greatly needed to combine these data to provide a comprehensive view of the underlying biology and human disease. An ideal method should be able to automatically extract relevant features by exploiting heterogeneous data across various modalities and dimensions to answer a specific biological or medical question. The key challenge in developing such methods is the design of a comprehensive model capable of harnessing noisy and high-dimensional datasets without much manual supervision. Recent advances in machine learning (ML) algorithms, especially deep learning, offer a unique opportunity to mine and provide a systematic understanding of massive heterogeneous biological datasets. In this chapter, we describe the principles of integrative genomic analysis and discuss existing ML methods. We provide examples of successful data integration in biology and medicine with a specific focus on omics (e.g., single-cell RNA-seq) and image data. Finally, we discuss current challenges in integrative methods for genomics and our perspective on the future development of the field.

Machine learning (ML) holds great promise for impacting healthcare delivery; however, to date most methods are tested in 'simulated' environments that cannot recapitulate factors influencing real-world clinical practice. We prospectively deployed and evaluated a random forest algorithm for therapeutic curative-intent radiation therapy (RT) treatment planning for prostate cancer in a blinded, head-to-head study with full integration into the clinical workflow. ML- and human-generated RT treatment plans were directly compared in a retrospective simulation with retesting (n = 50) and a prospective clinical deployment (n = 50) phase. Consistently throughout the study phases, treating physicians assessed ML- and human-generated RT treatment plans in a blinded manner following a priori defined standardized criteria and peer review processes, with the selected RT plan in the prospective phase delivered for patient treatment. Overall, 89% of ML-generated RT plans were considered clinically acceptable and 72% were selected over human-generated RT plans in head-to-head comparisons. RT planning using ML reduced the median time required for the entire RT planning process by 60.1% (118 to 47 h). While ML RT plan acceptability remained stable between the simulation and deployment phases (92 versus 86%), the number of ML RT plans selected for treatment was significantly reduced (83 versus 61%, respectively). These findings highlight that retrospective or simulated evaluation of ML methods, even under expert blinded review, may not be representative of algorithm acceptance in a real-world clinical setting when patient care is at stake.

We demonstrate operation of a compact, passively Q-switched Yb:YAG laser producing 10W average power at 1030nm in an 8.5kHz train of 1.2mJ, 1.6ns pulses. The laser is cw end-pumped with a fiber-coupled wavelength-stabilized 940nm diode. The pump light from the 105m diameter (0.1 NA) pigtail is collimated using a short focal length lens and directed into a 3mm long, 1mm thick 10 at.% Yb:YAG crystal. The saturable absorber is a Cr4+:YAG crystal with 76% unsaturated transmission. The laser’s slope efficiency is 52% and the optical-to-optical conversion efficiency is 33% at a pump power of 30W.

Quality assurance of data prior to use in automated pipelines and image analysis would assist in safeguarding against biases and incorrect interpretation of results. Automation of quality assurance steps would further improve robustness and efficiency of these methods, motivating widespread adoption of techniques. Previous work by our group demonstrated the ability of convolutional neural networks (CNN) to efficiently classify head and neck (H& N) computed-tomography (CT) images for the presence of dental artifacts (DA) that obscure visualization of structures and the accuracy of Hounsfield units. In this work we demonstrate the generalizability of our previous methodology by validating CNNs on six external datasets, and the potential benefits of transfer learning with fine-tuning on CNN performance. 2112 H& N CT images from seven institutions were scored as DA positive or negative. 153(8) images from a single institution were used to train three CNNs with resampling grid sizes of 64(3), 128(3) and 256(3). The remaining six external datasets were used in five-fold cross-validation with a data split of 20% training/fine-tuning and 80% validation. The three pre-trained models were each validated using the five-folds of the six external datasets. The pre-trained models also underwent transfer learning with fine-tuning using the 20% training/finetuning data, and validated using the corresponding validation datasets. The highest micro-averaged AUC for our pre-trained models across all external datasets occurred with a resampling grid of 256(3) (AUC = 0.91 +/- 0.01). Transfer learning with fine-tuning improved generalizability when utilizing a resampling grid of 256(3) to a micro-averaged AUC of 0.92 +/- 0.01. Despite these promising results, transfer learning did not improve AUC when utilizing small resampling grids or small datasets. Our work demonstrates the potential of our previously developed automated quality assurance methods to generalize to external datasets. Additionally, we showed that transfer learning with fine-tuning using small portions of external datasets can be used to fine-tune models for improved performance when large variations in images are present.

Enabling automated pipelines, image analysis and big data methodology in cancer clinics requires thorough understanding of the data. Automated quality assurance steps could improve the efficiency and robustness of these methods by verifying possible data biases. In particular, in head and neck (H&N) computed-tomography (CT) images, dental artifacts (DA) obscure visualization of structures and the accuracy of Hounsfield units; a challenge for image analysis tasks, including radiomics, where poor image quality can lead to systemic biases. In this work we analyze the performance of three-dimensional convolutional neural networks (CNN) trained to classify DA statuses. 1538 patient images were scored by a single observer as DA positive or negative. Stratified five-fold cross validation was performed to train and test CNNs using various isotropic resampling grids (64(3), 128(3) and 256(3)), with CNN depths designed to produce 32(3), 16(3), and 8(3) machine generated features. These parameters were selected to determine if more computationally efficient CNNs could be utilized to achieve the same performance. The area under the precision recall curve (PR-AUC) was used to assess CNN performance. The highest PR-AUC (0.92 +/- 0.03) was achieved with a CNN depth = 5, resampling grid = 256. The CNN performance with 256(3) resampling grid size is not significantly better than 64(3) and 128(3) after 20 epochs, which had PR-AUC = 0.89 +/- 0.03 (p -value = 0.28) and 0.91 +/- 0.02 (p -value = 0.93) at depths of 3 and 4, respectively. Our experiments demonstrate the potential to automate specific quality assurance tasks required for unbiased and robust automated pipeline and image analysis research. Additionally, we determined that there is an opportunity to simplify CNNs with smaller resampling grids to make the process more amenable to very large datasets that will be available in the future.

This study presents a comprehensive analysis of farmers' income in Sweden. The results indicate that farm households in Sweden do well from a standard-of-living perspective, but that farming is still a low-paid occupation from a return-on-skills perspective. Nevertheless, farm earnings increased faster over the study period than earnings in the general population, owing equally to higher farm earnings for operators and higher off-farm earnings for their spouse. Since few female spouses make farm earnings, evaluating farm household earnings from mainly a household perspective fails to acknowledge the individual careers of farmer and spouse.

Background The number of solid organ transplants in Canada has increased 33% over the past decade. Hospital readmissions are common within the first year after transplant and are linked to increased morbidity and mortality. Nearly half of these admissions to the hospital appear to be preventable. Mobile health (mHealth) technologies hold promise to reduce admission to the hospital and improve patient outcomes, as they allow real-time monitoring and timely clinical intervention. Objective This study aims to determine whether an innovative mHealth intervention can reduce hospital readmission and unscheduled visits to the emergency department or transplant clinic. Our second objective is to assess the use of clinical and continuous ambulatory physiologic data to develop machine learning algorithms to predict the risk of infection, organ rejection, and early mortality in adult heart, kidney, and liver transplant recipients. Methods Remote Mobile Outpatient Monitoring in Transplant (Reboot) 2.0 is a two-phased single-center study to be conducted at the University Health Network in Toronto, Canada. Phase one will consist of a 1-year concealed randomized controlled trial of 400 adult heart, kidney, and liver transplant recipients. Participants will be randomized to receive either personalized communication using an mHealth app in addition to standard of care phone communication (intervention group) or standard of care communication only (control group). In phase two, the prior collected data set will be used to develop machine learning algorithms to identify early markers of rejection, infection, and graft dysfunction posttransplantation. The primary outcome will be a composite of any unscheduled hospital admission, visits to the emergency department or transplant clinic, following discharge from the index admission. Secondary outcomes will include patient-reported outcomes using validated self-administered questionnaires, 1-year graft survival rate, 1-year patient survival rate, and the number of standard of care phone voice messages. Results At the time of this paper’s completion, no results are available. Conclusions Building from previous work, this project will aim to leverage an innovative mHealth app to improve outcomes and reduce hospital readmission in adult solid organ transplant recipients. Additionally, the development of machine learning algorithms to better predict adverse health outcomes will allow for personalized medicine to tailor clinician-patient interactions and mitigate the health care burden of a growing patient population. Trial Registration ClinicalTrials.gov NCT04721288; https://www.clinicaltrials.gov/ct2/show/NCT04721288 International Registered Report Identifier (IRRID) PRR1-10.2196/26816

Scientists conducting research in the field of LGBTQ+ Mental Health, like scientists in any other field, have the option of submitting their work to a more specialized or a more general interest jo...

Atlas-based machine learning (ML) for radiation therapy (RT) treatment planning is effective at tailoring dose distributions to account for unique patient anatomies by selecting the most appropriate patients from the training database (atlases) to inform dose prediction for new patients. However, variations in clinical practice between the training dataset and a new patient to be planned may impact ML performance by confounding atlas selection. In this study, we simulated various contouring practices in prostate cancer RT to investigate the impact of changing input data on atlas-based ML treatment planning. We generated 225 ML plans for nine bespoke contouring protocol scenarios (reduced target margins, modified organ-at-risk (OAR) definitions, and inclusion of optional OARs less represented in the training database) on 25 patient datasets by applying a single, previously trained and validated ML model for prostate cancer followed by dose mimicking to create a final deliverable plan. ML treatment plans for each scenario were compared to base ML treatment plans that followed a contouring protocol consistent with the model training data. ML performance was evaluated based on atlas distance metrics that are calculated during ML dose prediction. There were significant changes between atlases selected for the base ML treatment plans and treatment plans when planning target volume margins were reduced and/or optional OARs were included. The deliverability of ML predicted dose distributions based on gamma analysis between predicted and mimicked final deliverable dose showed significant differences for seven out of eight scenarios compared with the base ML treatment plans. Overall, there were small but statistically significant dosimetric changes in predicted and mimicked dose with addition of optional OAR contours. This work presents a framework for benchmarking and performance monitoring of ML treatment planning algorithms in the context of evolving clinical practices.

Automating the radiotherapy treatment planning process is a technically challenging problem. The majority of automated approaches have focused on customizing and inferring dose volume objectives to used in plan optimization. In this work we outline a multi-patient atlas-based dose prediction approach that learns to predict the dose-per-voxel for a novel patient directly from the computed tomography (CT) planning scan without the requirement of specifying any objectives. Our method learns to automatically select the most effective atlases for a novel patient, and then map the dose from those atlases onto the novel patient. We extend our previous work to include a conditional random field for the optimization of a joint distribution prior that matches the complementary goals of an accurately spatially distributed dose distribution while still adhering to the desired dose volume histograms. The resulting distribution can then be used for inverse-planning with a new spatial dose objective, or to create typical dose volume objectives for the canonical optimization pipeline. We investigated six treatment sites (633 patients for training and 113 patients for testing) and evaluated the mean absolute difference (MAD) in all DVHs for the clinical and predicted dose distribution. The results on average are favorable in comparison to our previous approach (1.91 vs 2.57). Comparing our method with and without atlas-selection further validates that atlas-selection improved dose prediction on average in Whole Breast (0.64 vs 1.59), Prostate (2.13 vs 4.07), and Rectum (1.46 vs 3.29) while it is less important in Breast Cavity (0.79 vs 0.92) and Lung (1.33 vs 1.27) for which there is high conformity and minimal dose shaping. In CNS Brain, atlas-selection has the potential to be impactful (3.65 vs 5.09), but selecting the ideal atlas is the most challenging., Comment: 18 pages, 4 figures, submitted to Physics in Medicine and Biology