Abstract Background Adaptive treatment strategies that can dynamically react to individual cancer progression can provide effective personalized care. Longitudinal multi-omics information, paired with an artificially intelligent clinical decision support system (AI-CDSS) can assist clinicians in determining optimal therapeutic options and treatment adaptations. However, AI-CDSS is not perfectly accurate, as such, clinicians’ over/under reliance on AI may lead to unintended consequences, ultimately failing to develop optimal strategies. To investigate such collaborative decision-making process, we conducted a Human-AI interaction case study on response-adaptive radiotherapy (RT). Methods We designed and conducted a two-phase study for two disease sites and two treatment modalities—adaptive RT for non-small cell lung cancer (NSCLC) and adaptive stereotactic body RT for hepatocellular carcinoma (HCC)—in which clinicians were asked to consider mid-treatment modification of the dose per fraction for a number of retrospective cancer patients without AI-support (Unassisted Phase) and with AI-assistance (AI-assisted Phase). The AI-CDSS graphically presented trade-offs in tumor control and the likelihood of toxicity to organs at risk, provided an optimal recommendation, and associated model uncertainties. In addition, we asked for clinicians’ decision confidence level and trust level in individual AI recommendations and encouraged them to provide written remarks. We enrolled 13 evaluators (radiation oncology physicians and residents) from two medical institutions located in two different states, out of which, 4 evaluators volunteered in both NSCLC and HCC studies, resulting in a total of 17 completed evaluations (9 NSCLC, and 8 HCC). To limit the evaluation time to under an hour, we selected 8 treated patients for NSCLC and 9 for HCC, resulting in a total of 144 sets of evaluations (72 from NSCLC and 72 from HCC). Evaluation for each patient consisted of 8 required inputs and 2 optional remarks, resulting in up to a total of 1440 data points. Results AI-assistance did not homogeneously influence all experts and clinical decisions. From NSCLC cohort, 41 (57%) decisions and from HCC cohort, 34 (47%) decisions were adjusted after AI assistance. Two evaluations (12%) from the NSCLC cohort had zero decision adjustments, while the remaining 15 (88%) evaluations resulted in at least two decision adjustments. Decision adjustment level positively correlated with dissimilarity in decision-making with AI [NSCLC: ρ = 0.53 ( p < 0.001); HCC: ρ = 0.60 ( p < 0.001)] indicating that evaluators adjusted their decision closer towards AI recommendation. Agreement with AI-recommendation positively correlated with AI Trust Level [NSCLC: ρ = 0.59 ( p < 0.001); HCC: ρ = 0.7 ( p < 0.001)] indicating that evaluators followed AI’s recommendation if they agreed with that recommendation. The correlation between decision confidence changes and decision adjustment level showed an opposite trend [NSCLC: ρ = −0.24 ( p = 0.045), HCC: ρ = 0.28 ( p = 0.017)] reflecting the difference in behavior due to underlying differences in disease type and treatment modality. Decision confidence positively correlated with the closeness of decisions to the standard of care (NSCLC: 2 Gy/fx; HCC: 10 Gy/fx) indicating that evaluators were generally more confident in prescribing dose fractionations more similar to those used in standard clinical practice. Inter-evaluator agreement increased with AI-assistance indicating that AI-assistance can decrease inter-physician variability. The majority of decisions were adjusted to achieve higher tumor control in NSCLC and lower normal tissue complications in HCC. Analysis of evaluators’ remarks indicated concerns for organs at risk and RT outcome estimates as important decision-making factors. Conclusions Human-AI interaction depends on the complex interrelationship between expert’s prior knowledge and preferences, patient’s state, disease site, treatment modality, model transparency, and AI’s learned behavior and biases. The collaborative decision-making process can be summarized as follows: (i) some clinicians may not believe in an AI system, completely disregarding its recommendation, (ii) some clinicians may believe in the AI system but will critically analyze its recommendations on a case-by-case basis; (iii) when a clinician finds that the AI recommendation indicates the possibility for better outcomes they will adjust their decisions accordingly; and (iv) When a clinician finds that the AI recommendation indicate a worse possible outcome they will disregard it and seek their own alternative approach.
Purpose: Overall low dose to lung for patients with locally advanced lung cancer with large PTVs is a concern, prompting investigation into the preferred arc geometry for lung VMAT. Methods: 8 subjects (6 males, 64±7 years, stage III/IV=5/3) were enrolled with a mean PTV size of 595±303cc. A standard dose scheme of 60Gy in 30 fractions was used for all subjects. For each subject, four VMAT plans with different combinations of the number of arcs (1 arc vs. two arcs) and the arc range (full vs. partial arc) were generated in Eclipse. Dosimetric characteristics for each plan were evaluated by PTV coverage and OAR sparing. PTV coverage was quantified by mean, max dose, homogeneity index HI5/95 defined as (D5%–D95%)/Mean_dose and conformity index (CI). OAR sparing was measured by mean, V10 and V20 for lung minus GTV, mean, max, V30 and V40 for heart and max dose for cord. For each metric abovementioned, a two‐way repeated ANOVA analysis was performed to evaluate whether the number of arcs and the arc range will significantly affect dose distribution of VMAT plans. Results: In general partial arc geometry provided better PTV coverage with significantly smaller HI5/95 (p<0.05) and higher mean dose (p=0.15) when keeping the total MU similar. For OAR sparing, partial arc geometry resulted in more dose sparing to the contralateral lung with significantly reduced mean lung dose (p=0.006). On the other hand, full arc geometry significantly reduced mean dose to the ipsilateral lung (p=0.02) as well as dose to the heart and the cord (p<0.05) especially with a two‐full arc geometry. Conclusion: Results of this study suggest that choice of an optimal VMAT geometry depends on clinical goals. For example, partial arc geometry should be preferred over full arc when considering overall low lung dose for patients with poor lung functions.
Emergence of big data analytics resource systems (BDARSs) as a part of routine practice in Radiation Oncology is on the horizon. Gradually, individual researchers, vendors, and professional societies are leading initiatives to create and demonstrate use of automated systems. What are the implications for design of clinical trials, as these systems emerge? Gold standard, randomized controlled trials (RCTs) have high internal validity for the patients and settings fitting constraints of the trial, but also have limitations including: reproducibility, generalizability to routine practice, infrequent external validation, selection bias, characterization of confounding factors, ethics, and use for rare events. BDARS present opportunities to augment and extend RCTs. Preliminary modeling using single- and muti-institutional BDARS may lead to better design and less cost. Standardizations in data elements, clinical processes, and nomenclatures used to decrease variability and increase veracity needed for automation and multi-institutional data pooling in BDARS also support ability to add clinical validation phases to clinical trial design and increase participation. However, volume and variety in BDARS present other technical, policy, and conceptual challenges including applicable statistical concepts, cloud-based technologies. In this summary, we will examine both the opportunities and the challenges for use of big data in design of clinical trials.
Shielded special nuclear material (SNM) detection is challenging because passive radiation emissions provide relatively little signal. Photon interrogation sources can induce (γ,n) and (γ,f) in fissionable material, providing a strong radiation signal for detection. In this work, we use an ion chamber and a sodium iodide detector to characterize the photons produced by a 9 MeV linear accelerator (linac) in our lab. An Exradin A12 ion chamber and steel attenuators are used to determine a half-value layer (HVL) for the linac to compare to the vendor data sheet. A 15.24-cm diameter, 15.24-cm height sodium iodide detector is then used to perform spectroscopy of the linac beam. To reduce pile-up, lead collimation is used to reduce the incident photon current and a thick aluminum attenuator was added to filter photons less than 4 MeV. Using the ion chamber, we measure a HVL of 2.76 cm with 95% confidence interval of [2.75, 2.79], which is in close agreement with our simulated HVL of 2.69 cm with 95% confidence interval of [2.67, 2.74]. The vendor data sheet documents 3.00 cm; the disagreement most likely stems from differences in the experimental setup. The measured linac spectrum appears to have an endpoint between 9 and 10 MeV, but the complex measured spectrum requires additional experiments and simulations to validate. Future work will investigate the neutron capture gamma-ray contribution from linac shielding to the detected spectrum, implement time-gating, and investigate the use of additional detector types to confirm the sodium iodide result.
<div>AbstractPurpose:<p>We hypothesized that optimizing the utility of stereotactic body radiotherapy (SBRT) based on the individual patient's probability for tumor control and risk of liver injury would decrease toxicity without sacrificing local control in patients with impaired liver function or tumors not amenable to thermal ablation.</p>Patients and Methods:<p>Patients with Child-Pugh (CP) A to B7 liver function with aggregate tumor size >3.5 cm, or CP ≥ B8 with any size tumor were prospectively enrolled on an Institutional Review Board–approved phase II clinical trial to undergo SBRT with baseline and midtreatment dose optimization using a quantitative, individualized utility-based analysis. Primary endpoints were change in CP score of ≥2 points within 6 months and local control. Protocol-treated patients were compared with patients receiving conventional SBRT at another cancer center using overlap weighting.</p>Results:<p>A total of 56 patients with 80 treated tumors were analyzed with a median follow-up of 11.2 months. Two-year cumulative incidence of local progression was 6.4% [95% confidence interval (CI, 2.4–13.4)]. Twenty-one percent of patients experienced treatment-related toxicity within 6 months, which is similar to the rate for SBRT in patients with CP A liver function. An analysis using overlap weighting revealed similar local control [HR, 0.69; 95% CI (0.25–1.91); <i>P</i> = 0.48] and decreased toxicity [OR, 0.26; 95% CI (0.07–0.99); <i>P</i> = 0.048] compared with conventional SBRT.</p>Conclusions:<p>Treatment of individuals with impaired liver function or tumors not amenable to thermal ablation with a treatment paradigm designed to optimize utility may decrease treatment-related toxicity while maintaining tumor control.</p></div>
TPS1098 Background: This randomized Phase II trial evaluates if stereotactic body radiotherapy (SBRT) and/or surgical resection (SR) of all metastatic sites in newly oligo-metastatic breast cancer who have received, or planned to receive, up to 12 months of first line systemic therapy without progression will significantly improve median progression free survival (PFS). If this aim is met, a phase III component evaluates if SBRT/SR improves 5 year OS. Secondary aims include local control in the metastases, new metastases, and technical quality. Translational primary endpoint is to determine whether < 5 CTCs is an independent prognostic marker for improved PFS and OS. Methods: A 2016 amendment expanded eligibility criteria to pathologically confirmed metastatic breast cancer to ≤ 2 sites and < 12 months of standard first line systemic therapy. Registration must occur <365 days of initial metastatic diagnosis. Primary disease must be controlled prior to registration. CNS metastases are ineligible. ER/PR and HER-2 neu is required on either the primary or metastatic site. Randomization is to standard systemic therapy with local radiotherapy/ surgery for palliation when necessary versus ablative therapy of all metastases with SBRT and/or SR. Statistics: For the phase IIR portion to detect a signal for improved median PFS from 10.5 months to 19 months with 95% power, 146 patients are required. The Phase III requires an additional 246 patients to definitively determine if ablative therapy improves 5-year overall survival from 28% to 42.5% (HR = 0.67), with 85% power and one-sided type I error of 0.025. For the translational research assuming a two-sided probability of type I error of 0.05, patients accrued in the Phase II-R/III portions will provide sufficient power of at least 91% to detect whether < 5 CTC's is prognostic. Contact Information: Protocol: CTSU member web site https://www.ctsu.org. Pt enrollment: OPEN at https://open.ctsu.org or the OPEN tab on CTSU member web site. Support: Supported by NRG Oncology grants U10CA180868 and U10CA180822 from the National Cancer Institute (NCI). Clinical trial information: NCI-2014-01810.
