The Quantitative Imaging Network The National Cancer Institute Then: 1939 And Now: 2016
Yüklə 4,95 Mb.
|
Acknowledgements We thank the ECOG-ACRIN and ACR core lab team members for the efforts in organizing and running the meeting, as well as assisting with this summary. Specifically Joy Brown, Donna Hartfeil, Jim Gimple, Dena Flamini, Lauren Uzdienski, and Adam Opanowski for their support, hard work, and good humor. We also thank the QIN and NCTN investigators for their thoughtful and vigorous participation. This effort was supported in part by NIH grant U01CA190254. MEETING AGENDA December 13, 2016 Objective: To facilitate interactions among thought leaders from the NCTN, QIN and related groups to discuss the quantitative imaging needs of oncologists for clinical trials and the tools imagers can offer to meet those needs. Outcome: Generate 4 – 6 ideas on how to develop prospective testing of quantitative imaging tools in national level clinical trials. 1. Introduction and overview – Robert Nordstrom and Mitch Schnall 2. Oncologist Speakers a. Systemic therapy– Alan Venook b. Locally target therapy– Mitch Machtay (NRG) c. Immunotherapy – Lei Zheng d. Precision oncology –Peter O’Dwyer 3. Quantitative Imaging speakers a. QIN – Larry Schwartz and Robert Nordstrom b. QIBA – Ed Jackson 4. QIN - NCTN quantitative imaging integration mechanisms a. ECOG-ACRIN – Paul Kinahan and Mark Rosen b. Alliance/SWOG – Michael Knopp 5. Breakout sessions (by imaging topic or oncology topics; this shows imaging topic focus)
6. Review breakouts, summary, and next steps – Lead by the leaders listed above NCI Overview and Possible Future Plans (Robert Nordstrom) From the QIN perspective, this was a very productive first meeting between NCTN representatives and QIN investigators. The structure of the meeting was appropriate for its content, and there was sufficient time devoted to the topics presented in this report. A follow-up meeting will be held as part of the upcoming QIN Annual Meeting in April at the NCI facility in Shady Grove, MD. This next meeting will review the highlights of the Philadelphia meeting and present a number of specific QIN tools considered advanced enough for NCTN consideration. The QIN investigators are in agreement with the timeline suggested in this report. Tool integration into clinical trials should begin as soon as possible, and it is understood that this would be a 1 to 2-year development effort. As a result, it is advantageous to begin the process as early as possible. This will involve prioritization of tools from the QIN perspective. This, then, will be a discussion topic at the upcoming QIN Annual Meeting. Larry Schwartz suggested QIN involvement in an upcoming Alliance meeting in Chicago. The is a good first-step in the integration process of tools into clinical trials. There were a number of positive discussions begun at the Philadelphia meeting, and there was understandably a focus on specific disease sites and types from oncology trialists and QIN members. While these discussions might help solve a problem faced by a QIN member in his or her research effort, they did not educate QIN members about tool translation. A follow-up meeting should include opportunities to highlight specific tools (or types of tools, e.g. segmentation) for critical discussion from both NCTN and QIN. Can we follow specific leads such as the breast volumetric tool from I-SPY? Additional discussions among the parties on topics such as industrial participation, the use of other NCI funding mechanisms (e.g. Academic/Industrial Partnerships), and possible public – private partnerships should also be on the docket. Breakout Group 1: Quantitative Response Assessment (L. Schwartz) Session Takeaways/Recommendations
Discussion Topics Colorectal Cancer
Coordination of Imaging, Clinical and Biopsy Data
Bone Lesions in Breast Cancer
RECIST Modifications
Earlier Identification of PD
Challenges for PET in CRC
Review Design
Software Needs
Other disease areas of interest
Obstacles
Breakout Group 2: Quantitative Biomarkers (D. Mankoff) The Quantitative Biomarkers breakout group covered several themes examining the viewpoint of the oncologist (what are trial needs) and the imager (what QI tools can meet these needs). In the final part of the breakout, the group pulled these perspectives into a discussion of both general approaches to QI biomarker testing, as well as specific possible trial ideas. A key insight in the discussion is that the integration of QI biomarkers tends to be somewhat disease-site and treatment specific. While there are some common themes, there is perhaps the impression that QI biomarker needs are somewhat specific to disease site and treatment approach. The discussion of specific disease sites focused largely on the interest and expertise of the breakout panel, with some overview of needs and goals as follows: 1. Breast – There is a need to assess residual disease post-therapy, especially neoadjuvant therapy, as well as early response. This led to a discussion focusing largely on breast imaging, especially MRI, and some volumetric methods for MRI, CT, and FDG-PET. Specific NRG trial ideas were noted. 2. Gyn - There was a desire to refine QI methods to assess response and residual disease for gyn cancer, especially ovarian cancer, akin to what is done in breast. This also focused on volume-based ideas and FDG-PET QI metrics for these clinical trials setting, with some proposed developing NRG trials that might work for tools focused in these areas. 3. Brain tumors – This discussion focused on target definition for radiotherapy and surgery, as well as assessing response to treatment and distinguishing residual tumor versus treatment effect. There was a strong desire to develop tools to delineate the tumor independent of BBB breakdown. The possible roles of MRS and tools for DCE- and DSC- MRI were highlighted, as were both NRG and ECOG-ACRIN trials where these might be applied. 4. H/N Cancer – This discussion also focused on target definition for radiotherapy and early response assessment. Metabolic volumes and standardized response assessment tools for FDG PET/CT were of interest, as were novel MR approaches as DWI. There was also interest in hypoxia imaging, given a resurgence of interest in hypoxia-directed therapy. While no specific prospective trials were discussed, it was noted that the FDG PET/CT image dataset from ACRIN 6685 would be ideal for tools for target definition and metabolic volume measures. Specific trial ideas are noted in Section C. The breakout group closed by suggesting that interested QIN members in the areas noted above might reach out to disease-site leaders (including those in the breakout group) in the relevant NCTN group and might present their tools in future group meetings. In addition, the discussion targeted high-value datasets from ongoing trial (e.g., ACRIN6685) to test QIN tools.
Discussion notes are provided by disease site, largely representing the interests of oncologists in the breakout groups. Ovarian cancer 1. Biggest need in GYN studies is ovary drug approval and defining what is the actual endpoint. 2. The pCR rates are in single digits in trials – need targeted therapeutic but the problem is the data from phase 3 show 50% false negative rate of complete response; microscopic disease was found when patient went to surgery. Could we have an imaging correlate with this to better define pCR (identifying residual viable disease and pointing the surgeons in the right direction), as pathologic CR has currently a low correlation with imaging 3. We may also need to look at how pCR is defined in ovarian cancer; there is a lot of imaging data from neoadjuvant data in breast but not as much in ovary 4. NRG (especially GOG component) trial interest
1. How do you predict who the patients are that will do well – the mammography imaging we have now is not highly predictive; 2. NRG is developing a study using core needle biopsy along with triple imaging to predict complete responders. One of the problems in breast is post neoadjuvant data in imaging; this led to suggesting the trial with triple imaging. 3. One issue is nodal response, if we are trying to eliminate surgery in the breast but still need to do a sentinel node biopsy: is there an imaging test to be predictive of the nodes so as to save surgery for patients. 4. There is some desire to make neoadjuvant treatment for breast a non-surgical endpoint; pCR in breast right now is no tumor in breast or no tumor in nodes: can we come up with a marker for this and an imaging correlate to help with microscopic disease. In lymphoma – the negative FDG-PET has been predictive, could this work in breast, even if FDG-PET misses small-volume disease, knowing that other treatments (radiation, systemic therapy) have been effective with microscopic disease.
1. MRI visualizes the blood brain barrier; this works if the barrier is broken down by the tumors (high grade lesions), but is not as accurate without a breakdown in the barrier. 2. Challenges in brain tumors involve assessment when radiation therapy is involved, defining the active tumor versus radiation effects and response. This is both for up-front management and identifying residual or recurrent tumor (versus treatment effect) post-therapy. 3. We also need an early marker for response to determine patient management - something that measures the tumor itself - i.e., something that can image more than just the blood brain barrier. Head and Neck Cancers 1. Radiation Specialist: There is interest in using quantitative imaging tool and identify targets and treat those targets (early data in lung and head and neck; could also apply to brain) 2. From a trial perspective, there are enormous opportunities to look at tumor targeting based on quantitative imaging. 3. H/N Cancers generate an interest in hypoxia imaging to guide a renewed interest in hypoxia-targeted therapy. 4. Can quantitative imaging of H/N cancer help guide use of combined, multimodal therapy - including molecularly targeted therapy (e.g., EGFR-targeted agents) and immunotherapy.
1. What tools does Radiology have to offer based on the oncology discussion:
Check on RTOG 0522 and the # of cases that have PET images for evaluation. 2. Immunotherapy: If we have a clinical trial that is using PD1 inhibitors – is there a way to do the data collection and imaging to correlate with the biomarker that is being looked at
3. In terms of approaches to breast MRI (for example, high spatial resolution – low temporal resolution):
4. Brain tumor DSC MRI has shown predictive and we are using it in ECOG-ACRIN EAF151 (open in first half of 2017). These seem to be low-hanging fruit for groups working in this area. Could spectroscopic aims be added?
5. Head and neck cancer image analysis development ideas:
C. Summary discussion – bring oncologist and imager perspective together 1. What can we do to work with QIN sites to translate tools into clinical trials?
2. Implement an image analysis tool that helps to refine analysis and possibly reduce the number of needed patients if the quantitative accuracy of data can be made more consistent. 3. Develop integrated markers to implement (target) specific disease-site trials
4. Need to find other pathways to get live data – embed the collection or information in a study in order to collect it for future. 5. The clinicians are not aware of the tools that QIN has or is working on, that will be available for them to help them. There has to be a better way of "marketing" the tools to the clinicians. 6. There needs to be more cross talk between the Clinicians and the investigators on the tools that are needed by the clinicians and what can be easily validated/ developed / modified and made available to them.
NCI-MATCH trial affords an opportunity to assemble a database of radiologic imaging with corresponding genomic profiling in 6000 subjects. Our discussion aims to define the scientific, technical, regulatory and political objectives for this endeavor. Oncologist Perspective: Peter O’Dwyer, ECOG-ACRIN “Precision medicine is getting the right drug to the right patient at the right dose.” Challenge is that cancer is heterogeneous. This heterogeneity is likely a moving target and is thought to be the basis for eventual progression through chemotherapy. We need methods to measure and predict tumor heterogeneity. Genomic sequencing continues to become cheaper but repeated tumor biopsy is invasive and not the solution. Oncology is looking to radiology to develop tools to describe propensity of tumor for heterogeneity. Biomarker Science Perspective: Susanna Lee, ECOG-ACRIN Oncologists are looking for the “ready to roll out” biomarkers --- imaging tools that are robust and standardized to use in clinical trials. Quantitative imaging methods currently available within the QIN labs and other centers have, for the most part, not matured enough for adoption in clinical trials. Purpose of MATCH imaging database is to provide a repository with which the QIN labs can continue to develop their methods. ECOG-ACRIN is working to design a database framework that will allow for meaningful integration of the imaging, clinical and genomic data, and enable ready access to imaging labs to carry out hypothesis generating research. Image Analytics Perspective: Andrey Fedorov and Yoga Balagurunathan, QIN MATCH is collecting standard of care imaging not acquired under a pre-defined protocol. It is not annotated. This represents non-controlled and non-normalized data and applying quantitative methods to them to yield meaningful results will be challenging. Need for control regions and phantoms would be essential to control imaging system variability across centers and diversity of scanners. Integrating the images with the reports, biopsy imaging, NGS results and clinical diagnoses is necessary for this database to be useful. The cohort is heterogeneous with respect to tumor type and driver mutations. Thoughtful and detailed curation is essential. Ensuring access and enabling data sharing is important. Quantitative analysis of digital pathology images is a very promising venue, since pathology slides will be collected in a centralized location and QIN is involved is interested both in digital pathology (pathomics) studies, and correlation of pathology to radiological and genomic findings. Imaging Standardization Perspective: Ed Jackson, QIBA Imaging heterogeneity represents a major challenge for the QIN labs and for quantitative imaging biomarker development in general. It will again prove to be a hurdle when analyzing the MATCH database. QIBA approach is to identify sources of imaging variability and specify potential solutions in the form of profiles. Investment into testing these profiles must be made before dissemination to vendors and users. Compliance Perspective: Betsy Hsu, NCI MATCH patients have consented to image collection. Image analytics will require IRB approval as a retrospective medical records review. Conclusion Challenges and goals for informatics and database development for precision medicine are:
Breakout Group 4: Image Accrual and Curation (P. Kinahan, M. Rosen) Overview This breakout group focused on the central questions of the accrual and quality of images and associated data. The challenges were covered in earlier presentations at the QIN-NCTN workshop by Jackson, Kinahan, Knopp, and Rosen and were reviewed in more detail in this breakout session. This group started discussion from the viewpoint that there is a clear value in the accrual of prospective imaging data and other clinical data for new classes of therapies, for example immunotherapies and combination therapies. Due to the choice of topic, this breakout session did not focus on specific ongoing or planned clinical trials that might include a quantitative imaging endpoint employing a QIN tool or method. Rather, this discussion centered around the interactions between such clinical trials and QIN tools and methods with that goal of ensuring proper planning and increasing the likelihood of success of a QIN tool or method in a clinical trial using a quantitative imaging endpoint. With this goal in mind, the discussions of this breakout group were clustered around four topics:
For each of the above topics the discussions are summarized into three subgroups: Benefits, Challenges, and Potential solutions.
The NCI Community Oncology Research Program (NCORP) conducts multi-site cancer clinical trials and studies in diverse populations in community-based healthcare systems. As was well described in the presentation by Michael Knopp, image and data collection from NCORP sites has unique challenges. For example, two-thirds of sites only participate in one or two trials and sites typically do not have the expertise of NCTN imaging sites. Benefits: If appropriately collected, large amounts of data could be available from NCORP sites to power studies. This includes images and access to modern genetic/pathology data. Challenges: The collection and submission of images and other clinical data is onerous for NCORP sites given the lack of redundancy in staffing and limited expertise. For the same reasons, implementation of necessary QA/QC procedures and the standardization of image acquisition and reconstruction is also problematic for NCORP sites. As noted above 66% of NCORP sites participate in only one or two trials. Thus, there is no opportunity for the efficiency of scale that NCTN sites can achieve when participating in multiple studies. As the NCORP sites have few personnel and NCI trial are typically underfunded, any extra effort or cost requires careful justification locally. Potential Solutions: A central theme in the breakout session that is perhaps most relevant here was that NCORP-based studies need analytic methods and tools from QIN that are robust to variations in data quality. In other words, if QIN methods and tools are developed based on the assumption that sites follow rigorous QA/QC procedures and standardized of image acquisition and reconstruction procedures, they will likely not be useful with data collected from NCORP sites. Conversely, any QIN methods or tools that have a low barrier to implement and also robust to variations in data quality will allow access to the large amounts of image and clinical data collected from NCORP sites.
The 30 NCI National Clinical Trials Network (NCTN) imaging sites, in comparison to NCORP sites, are larger programs primarily based at NCI-designated Cancer Centers. Benefits: NCTN site have multiple trials and can take advantage of economies of scale to ensure QA/QC procedures, standardization of image acquisition and reconstruction procedures, and data transmittal. Higher quality data (on average) and better access to clinical data. In addition, trial PIs are generally located at NCTN sites, helping to form more of an 'ecosystem' where trial information is more widely available. Challenges: Even though NCTN sites are larger, any extra effort or cost still requires justification. This, QA/QC and standardization of image acquisition is problematic even for NCTN sites. While NCTN sites have large repositories of imaging and clinical data, privacy and regulatory concerns, whether real or perceived, limit access to data. Clinical trials typically take multiple years, with no access to data until a year after the conclusion of accrual or publication of the results. This is often stipulated in the NCI award letters, for example the Match Imaging trial. Potential Solutions: Similar to the discussion above for NCORP sites, we need analytic methods and tools from the QIN that are robust to the data quality that will be collected from NCTN sites. I.e. not methods that are only effective with ‘best possible’ quality data. QIBA Profiles can be used to reduce the bias and variance of imaging biomarkers. To provide earlier access to imaging and clinical data three potential solutions were proposed: Sites or QIN investigators can use existing specimen biobanks as models. These are sometimes funded by NIH or locally by larger institutions. Second: Strengthen the ‘ecosystems’ where there is easier access to data between collaborative groups, even for ongoing trials. Third: Determine/publish/advocate systems where trial data is (albeit probably without outcomes) is available even before the conclusion of the trial.
Benefits: There is a lot of high quality imaging and clinical data already available. Typically, this imaging and clinical data is well-curated and often has outcomes available. Challenges: As always, any extra effort/cost requires justification. In addition, older data may be less relevant, for both images and therapies. Potential Solutions: The ECOG-ACRIN QIN team is already in the process of making legacy ACRIN imaging trial data available to QIN members through the Cancer Imaging Archive (TCIA) system. In addition, the Quality Assurance Review Center (QARC) has longstanding history of imaging and radiotherapy trial process development and improvement. QARC has one of the largest and most diverse multisite oncology archives. While the ACRIN-TCIA access in known by many QIN members, the QARK data may be less so. Thus it makes sense to provide a virtual or on-line tour of the QARK data archives to QIN members, either as a webinar or at a QIN annual meeting.
Many QIN members are imaging scientists who are unfamiliar with clinical trials and specifically oncology clinical trials. Benefits: Helping QIN members less experienced in oncology clinical trials gain access to new or existing trials will aid in the development of tools and methods that improve the efficacy or efficiency of clinical trials of new cancer therapies. In addition, linking imaging experts to therapy experts can increase participation and support from pharma, scanner manufacturers, iCROs, and regulatory bodies etc. Challenges: Imaging scientists and clinical trialists tend to work in very different fields of study. In addition, clinical trialists working in therapeutic developments may be at best neutral towards including advanced imaging methods that may help their trials, due to complexity, time, and cost. While there are Finally, while there are centralized databases, e.g. TCIA, ACR-TRIAD, QARK, there is no centralized list of databases for imaging researchers. Potential Solutions: As noted above, it makes sense to provide a virtual or on-line tours of the QARK data archives to QIN members, either as a webinar or at a QIN annual meeting. It would also be helpful to provide a list of centralized databases to the QIN, e.g. TCIA, ACR-TRIAD, QARK. In addition, QIN members should make a point of attending NCTN group meetings. While the ECOG-ACRIN QIN team has been doing this on conference calls and the last two QIN annual meetings, these encouragements should continue. In a reciprocal fashion, clinical trial PIs should be invited to conference calls and to the QIN annual meeting, following the program initiated by Larry Schwarz. Finally, workshops and planning meetings such as this one are clearly needed to bring together QIN members with thought leaders from the NCTN and related groups for roundtable discussions on what oncologists need for quantitative imaging with their oncology trials and what imagers can offer to improve the efficacy of these trials.
This breakout group discussion identified several benefits and challenges in image accrual and curation from the viewpoint of including advanced imaging methods as tools or methods that improve the efficacy or efficiency of clinical trials of new cancer therapies. While no potential trials using advanced imaging methods were proposed by this group, the discussion did identify central challenges to this process. However, the discussion identified potential solutions and directions that can enable the incorporation of tools or methods developed by the QIN in NCTN trials are now proposed for each of the topics as described above. WORKING NOTES FROM PLANNING MEEETING Breakout Group 1: Quantitative response assessment (L. Schwartz)
Breakout Group 2: Quantitative biomarkers (D. Mankoff)
Breakout Group 3: Informatics/precision metrics (S. Lee, A. Fedorov)
Breakout Group 4: Image accrual and curation for analysis (P. Kinahan, M. Rosen)
Summary Themes for all 4 breakout groups:
Open discussion:
Attendees 
Appendix 2: Agenda for the NPI Workshop Strategies for Improving Early Detection of Cancer and Response to Therapies through Imaging Technologies April 12, 2017 Cambria Hotel, Rockville, MD NPI Workshop on Strategies for Improving Early Detection of Cancer and Response to Therapies through Imaging Technologies April 12th, Cambria Hotel, Rockville MD TENTATIVE SCHEDULE 9:00 – 9:15 Welcome, National Photonics Initiative overview and review of charge 9:15-10:15 Session 1: The National Cancer Moonshot Initiative and Complementary initiatives from various federal agencies Organizer Eva M. Sevick-Muraca, Professor and Chair, University of Texas Health Science Center
10:15 – 11:15 Session 2: Medical imaging and evidence development: IT, protocol standards, multi-disciplinary associations, and technology assessment Organizer: Maryellen Giger, Professor, University of Chicago (a) Quantitative Imaging - Ed Jackson (b) Distributed Trials and Cloud Computing – Claudia Henschke and David Yankelevitz (c) Imaging Genomics and Deep Learning – Maryellen Giger (d) Technology Assessment – Berkman Sahiner 11:15-11:30 Break 11:30 -12:30 Session 3: Multi-Modal Diagnostics for Predictive Modelling via High-Performance Computing; Decision support based on clinical evidence from community oncology practice Organizer: Richard Franks, Chief Medical Officer, Siemens Healthineers USA (a) Innovations in Screening and Surveillance for Prostate Cancer - Minhaj Siddiqqi (b) Innovations in Diagnosis, Staging, Choice of Treatment, and Recurrence - Peter Choyke (c) Innovations in Quantitative Imaging - Rick Avila (d) Deep Learning to Model Care Pathways - Maciej Mazurowski; 12:30-12:45 Break and lunch service (working lunch) 12:45- 2:30 Session 4: Discussion, Summary of Recommendations and Next Steps Discussion Lead: James Mushine Acting Dean and Vice President for Research, Rush University Medical Center Yüklə 4,95 Mb. Dostları ilə paylaş:
|