EPIDEEP: Using a Deep Learning Model to Predict In Vivo Electronic Portal Imaging Device (EPID) Transit Images

PURPOSE/OBJECTIVE(S)

To create a deep-learning model to predict in-vivo electronic portal imaging device (EPID) transit images for IMRT treatments. This model was created to predict in-vivo images to identify machine and patient-related errors that occur during beam delivery and are undetectable with current QA approaches. The deep-learning model can make image predictions much faster than Monte Carlo approaches, making image prediction feasible for application in online adaptive radiotherapy. Additionally, the model does not rely on any proprietary information and can be easily utilized by other clinics.

MATERIALS/METHODS

Our approach separates the primary and scatter components of in-vivo transit images. The attenuation of primary radiation reaching the EPID panel is modeled analytically, using attenuation measurements from phantoms of known thicknesses. The scatter component is estimated using a convolutional neural network (CNN). The CNN training uses information from the on-treatment cone-beam CTs (CBCTs), and a pretreatment EPID image with no patient in the beam. We acquired 193 IMRT fields/images from 118 patients previously treated on the Varian Halcyon. Treatment sites included the pelvis, abdomen, lungs, and extremities. CBCTs were collected immediately before treatment, to provide an accurate representation of the anatomy. A 3-channel input image was used, consisting of the pretreatment EPID image, a ray tracing projection through the CBCT to the EPID panel, and a projection to isocenter. Model training:validation:test set ratios were 133:20:40 images. The primary and scatter components are added together to give the predicted transit image. Prediction accuracy was assessed by comparing model-predicted and measured in-vivo EPID images with a 3%/3mm and 5%/3mm gamma pass rate.

RESULTS

The gamma pass rate for the patients in the training:validation:test was 91.5%:90.4%:92.1% for 3%/3mm and 96.7%:96.6%:97.0% for 5%/3mm. The model can make image predictions in 20 milliseconds. The poor passing rates of some images may be due to CBCT artifacts and patient motion that occurs between the time of CBCT and treatment.

CONCLUSION

This model can predict in-vivo EPID images with an average gamma pass rate greater than 90%. Image predictions from this model can be used to detect in-vivo treatment errors and changes in patient anatomy, providing an additional layer of patient-specific quality assurance. The speed of image predictions is 20 milliseconds, making use feasible for online adaptive treatments, which currently do not utilize patient-specific measurements of the delivered radiation. Upcoming studies will assess the model's ability in detecting clinically relevant errors and changes in patient anatomy that can adversely affect treatment. Future goals include acquiring more data to further improve the model and extending the model to make predictions for VMAT treatments.

O-125 Application of artificial intelligence using big data to devise and train a machine learning model on over 63,000 human embryos to automate time-lapse embryo annotation

Study question

Can a machine learning (ML) model, developed using modern neural network architecture produce comparable annotation data; utilisable for algorithmic outcome prediction, to manual time-lapse annotations?

Summary answer

The model automatically annotated unseen embryos with comparable results to manual methods, generating morphokinetic data to enable comparably predictive outputs from an embryo selection algorithm.

What is known already

The application of artificial intelligence across healthcare industries, including fertility, is increasing. Several ML models are available that seek to generate or analyse embryo images and morphokinetic data, and to determine embryo viability potential. Along with photographic images, the use of time-lapse in IVF laboratories has amassed numeric data, resulting predominantly from annotated manual assessment of images over time. Embryo annotation practice is variable in quality, can be subjective and is time-consuming; commonly taking several minutes per embryo. The development of rapid, accurate automatic annotation would represent a significant time-saving as well as an increase in reproducibility and accuracy.

Study design, size, duration

Multicentre quality assured annotation data from 63,383 time-lapse monitored embryos (EmbryoScope®), comprising over 400 million individual images, were used to train a ML model to automatically generate morphokinetic annotations. Data was derived from 8 UK clinics within a cohesive group between 2012-2021. Accuracy was assessed using 900 unseen embryos (with live birth outcome) by comparing the output of an established in-house, prospectively validated embryo selection model when the input was either ML-automated, or manual annotations.

Participants/materials, setting, methods

Multi-focal plane images were processed on the Azure cloud (Microsoft) and resampled to 300x300 pixels. A Laplacian-based focal stacking algorithm merged frames into a single image. The model consisted of an EfficientNetB4 Convolutional Neural Network classifier to extract features and classify the stage of embryo images. A Temporal Convolutional Network  interpreted a time-series of image features; producing annotations from pronuclear fading through to blastocyst. Soft localisation loss function used QA data to integrate annotation subjectivities.

Main results and the role of chance

The ML model rapidly and automatically generated annotations. Efficacy and comparability of the ML model to automate reliable, utilisable annotations was demonstrated by comparison with manual annotation data and the ML model’s ability to auto-generate annotations which could be used to predict live birth by providing annotation data to an established, validated in house embryo selection model. Live birth-predictive capability was measured, and benchmarked against manual annotation, using the area under the receiver operating characteristic curve (AUC).

When tested on time-lapse images, collected from pronuclear fading to full blastulation, representing 900 previously unseen, transferred blastocysts where live birth outcomes were blinded, the in-house developed auto-annotation ML model resulted in an AUC of 0.686 compared with 0.661 for manual annotations, for live birth prediction.

Auto annotation using the developed model took only milliseconds to complete per embryo. The developed auto-annotation model, built and tested on large data, is considered suitable for productionisation with the aim of being validated and integrated into an application to support IVF laboratory practice.

Limitations, reasons for caution

Whilst this model was trained to recognise key morphokinetic events, there are other morphokinetic variables that may be useful in the prediction of live birth and further improve embryo selection, or deselection, ability. Akin to manual interpretation, some embryos may fail to be annotated or need second opinion.

Wider implications of the findings

There is increasing evidence supporting the application of ML to utilise big data from time-lapse imaging and fertility care generally. Whilst promising benefits to IVF clinics and patients, responsible use of data is required alongside large high-quality datasets, and rigorous validation, to ensure safe and robust applications.

Trial registration number

N/A

O-306 LINZAGOLIX FOR ENDOMETRIOSIS-ASSOCIATED PAIN: SAFETY RESULTS FROM EDELWEISS 3, A PHASE 3, RANDOMIZED, DOUBLE-BLIND, PLACEBO-CONTROLLED TRIAL

Study question

Is once-daily linzagolix treatment for women with moderate to severe endometriosis-associated pain (EAP) safe for use for up to 6 months of treatment?

Summary answer

Both doses of linzagolix were well-tolerated with minimal BMD decrease and few TEAEs >5% in either linzagolix arm.

What is known already

Linzagolix (LGX) is an investigational once-daily oral GnRH receptor antagonist that reduces serum estradiol in a dose-dependent manner and is being developed in two dosages for the treatment of endometriosis-associated pain (EAP): 75 mg, and 200 mg dose with hormonal add-back therapy (ABT). 

Study design, size, duration

EDELWEISS 3 is a randomized, double-blind, placebo-controlled, multicenter Phase 3 trial of linzagolix in women with moderate to severe EAP. The trial includes 3 treatment arms: 75 mg LGX, 200 mg LGX with ABT (E2 1 mg/ NETA 0.5 mg), or placebo. Here we present safety results up to 6 months (24 weeks) of treatment.

Participants/materials, setting, methods

Eligible reproductive-aged women with moderate-to-severe EAP were randomized and treated (n = 484) for 6 months with 75 mg LGX, 200 mg LGX with ABT (E2 1 mg/ NETA 0.5 mg), or placebo. Safety and tolerability objectives reported here include 6-month results for treatment emergent adverse events (TEAEs), assessment of mean percent change from baseline (CfB) in lumbar spine (LS) bone mineral density (BMD) and Z-scores.

The safety analysis set included 484 subjects across the 3 treatment groups.

Main results and the role of chance

The overall incidence of TEAEs was similar between the placebo and LGX 75 mg group (46.9%) and slightly higher (56.8%) in the LGX 200 mg + ABT group. There were few (3) serious TEAEs, and none were related to LGX. TEAEs that occurred in over 5% of patients in either active treatment arm included headache (10.5%, 8.1%, and 8.0%), hot flush (6.8%, 7.5%, and 2.5%), and fatigue (6.8%, 3.8%, and 2.5%) for the 200 mg LGX with ABT, 75 mg LGX, and placebo groups, respectively. Mean percent CfB (95% CI) in LS BMD was -0.79% (-1.15,  -0.43%), -0.89% (-1.31, -0.47%), and +0.78% (0.41, 1.15%) for the 200 mg LGX with ABT, 75 mg LGX, and placebo groups, respectively. Z-scores at 6 months remained within the same range as baseline in all groups.

Limitations, reasons for caution

Additional efficacy and safety results from the trial's 24 weeks (6 mo) extension phase are pending. (Edelweiss 6 protocol: NCT04335591)

Wider implications of the findings

These results support further development of ABT and non-ABT doses of linzagolix that have potential to provide flexibility and choice for women seeking treatment for EAP. A non-ABT option is important for women who have a contraindication to, are at increased risk for complications, or prefer not to use ABT.

Trial registration number

ClinicalTrials.gov: NCT02778399

Introduction to the complex intervention concept and the research perspectives

 

Few interventions are truly simple; complexity can arise from various sources, e.g.:

Aspects of the intervention itself, such as the number of intervention components, the number of groups targeted, how dynamic the intervention is permitted to be. The context in which the intervention is developed and delivered, such as the social, political, economic, and geographical context.

Complex intervention research should be approached with an awareness of these sources of complexity. Systems thinking can be helpful to understanding the dynamic interaction between interventions and their context. This presentation will introduce concepts of complex adaptive systems, e.g. feedback loops, adaptation, emergence, that should be considered when developing and evaluating complex interventions. It will then introduce participants to the research perspectives set out in the new framework: efficacy, effectiveness, theory-based, and systems perspectives. Each perspective is associated with a different type of research question, and therefore appropriate in different circumstances. The presentation will provide information to support participants to consider the research perspective(s) most suited to the research challenge that they are aiming to address.

Main messages

There are multiple sources of complexity, each of which can affect how the intervention works or contributes to change. Complex intervention research can take an efficacy, effectiveness, theory-based, or systems perspective, the choice of which is based on what is known already and what further evidence would be most useful.

The new Framework and the Core Elements of complex intervention research

 

The new framework will be presented. The main phases of intervention research: development or identification, feasibility, evaluation and implementation are connected to 6 core elements:

Context: any feature of the circumstances in which an intervention is conceived, developed, evaluated and implemented Programme theory: how an intervention is expected to lead to its effects and under what conditions. Programme theory should be tested and refined at all stages and used to guide the identification of uncertainties and research questions Stakeholders: those who are targeted by the intervention, involved in its development or delivery, or more broadly those whose personal or professional interests are affected, that is who have a stake in the topic. This includes patients, the public, and professionals Refinement: the process of ‘fine tuning' or making changes to the intervention once a preliminary version has been developed Uncertainties: identifying key uncertainties that exist given what is already known and what the programme theory, researchers and stakeholders identify as being most important to find out. These judgements inform the framing of research questions that, in turn, govern research perspective choice Economic considerations: exploring the comparative resource and outcome consequences of the interventions for those people and organisations affected

The presentation will discuss how to use the framework, highlighting that complex intervention research can be an iterative process. Repeating of phases is preferable to automatic progression to the next phase if uncertainties remain unresolved.

Main messages

Complex intervention research may begin at any phase, depending on what is appropriate for the intervention in question, and does not necessarily move sequentially through the phases. The core elements should be considered early and revisited continually throughout, as this will make it most likely that the intervention will be implementable in practice.