Radiology 2040

医学 百年 专业 放射科 核医学 病理
作者
James A. Brink,Hedvig Hricak
出处
期刊:Radiology [Radiological Society of North America]
卷期号:306 (1): 69-72 被引量:9
标识
DOI:10.1148/radiol.222594
摘要

HomeRadiologyVol. 306, No. 1 PreviousNext Reviews and CommentaryFree AccessEditorial–Centennial ContentRadiology 2040James A. Brink, Hedvig Hricak James A. Brink, Hedvig Hricak Author AffiliationsFrom the Department of Radiology, Massachusetts General Hospital, Brigham and Women’s Hospital, Boston, Mass (J.A.B.); and Department of Radiology, Memorial Sloan Kettering Cancer Center, 1275 York Ave, Ste H-704, New York, NY 10065 (H.H.).Address correspondence to H.H. (email: [email protected]).James A. BrinkHedvig Hricak Published Online:Dec 19 2022https://doi.org/10.1148/radiol.222594MoreSectionsPDF ToolsImage ViewerAdd to favoritesCiteTrack CitationsPermissionsReprints ShareShare onFacebookTwitterLinked In Download as PowerPointOpen in Image Viewer A translation of this article in Spanish is available in the supplement.Una traducción de este artículo en español está disponible en el suplemento.IntroductionNothing is permanent except change.– Heraclitus, around 500 BCEHistorically, radiology has been a specialty of innovation. Radiology has embraced new technologies and recognized their potential well before their clinical value has been widely accepted. We have been fearless in change and adaptation. Over the past 4 decades alone, we have repeatedly expanded and reshaped our field to accommodate revolutionary technological advances. We adopted cross-sectional anatomic imaging with CT and MRI, which required adaptation to much greater levels of detail in image interpretation. We introduced various forms of functional and molecular imaging, which required us to incorporate biology into our knowledge and image interpretation. We facilitated tremendous growth in applications of interventional radiology, incorporating clinics and clinicians with admitting privileges into many of our departments. Finally, most recently, we introduced marked advances in radiotheranostics, which combines targeted molecular imaging with radionuclide therapies.Today, we are in some ways the victims of our own success, as the demand for imaging, especially in oncology, keeps increasing. We now face large volumes of examinations, exponential increases in the number of images generated by each examination, and the demand for 24/7 coverage. In an era of widespread physician burnout, our specialty is among the most affected.The present is not sustainable, and yet again, radiology is at a crossroads. The next several decades are sure to bring seismic organizational and operational changes that will affect all medical specialties, including radiology. Artificial intelligence (AI) is emerging as an extremely powerful force in medicine and particularly in radiology—one that could help streamline our workflow and reduce our workload while also posing challenges to our relevance. Moreover, the pace of technological change is constantly accelerating. Many operational and technological developments that will affect radiology are being driven by external forces beyond our control. Imaging will still play a critical role in diagnosis and minimally invasive intervention. But, as radiologists, our focus must be laser-sharp on ways in which we can add value to the health care continuum, especially given the rise of non–image-based precision diagnostics and minimally invasive interventional techniques certain to occur in other specialties. Radiologists whose activities are limited to image interpretation will not only become a commodity but, eventually, may become obsolete. It is time to actively search out and demonstrate new means through which we can contribute beyond image interpretation. If we approach this challenge in the right way, then the specialty will continue to prosper. But the specialty will have to change considerably.At the most basic level, radiologists’ ability to remain clinically relevant will depend on our becoming and staying part of the clinical team. The ability to offer therapy will also be essential for strengthening our clinical impact. Thus, we will need to cultivate interventional radiology and radiotheranostics extensively. Furthermore, it will be critical for us to partner with other disciplines—in medicine, data science, engineering, and other areas—to perform clinically relevant research and carve out paths of innovation and expertise unique to our specialty. This editorial describes a variety of anticipated changes in the science and practice of radiology (Table), some of which will appear almost inevitably and some of which the imaging community will only be able to achieve through vision and intense determination.What to Know to Ensure Radiology Is Thriving in 2040Radiology Workflow and TeleradiologyIncreasingly, computation and data analytics will support imaging and other diagnostics, as well as connectedness and telehealth (1). Current workflows are largely driven by patient location. Worklists are organized according to whether patients are inpatients, outpatients, or admitted to the emergency department. But we will see the distinction between these three patient locations blurred as more and more medical care is delivered at home. Home hospital programs are on the rise. If inpatient care can be delivered at home, it is likely that we will see more urgent and acute care being delivered at home as well. Thus, our workflows will need to be driven by disease focus rather than patient location.Just as other medical care will be delivered at home, imaging technologies will increasingly be brought to the patient for self-examination. AI-powered autonomous or semiautonomous US using low-cost transducers driven by smartphone technology will enable patients to perform simple US data acquisition on their own, with images being reconstructed automatically (2). In the future, other examinations may be performed locally, including with portable equipment for radiography, CT (motionless), and MRI (low field strength) (3,4).Home reading and teleradiology will continue to rise due to the growing demand for worker flexibility and the need for continuous coverage as well as outreach beyond urban centers. However, radiologists of the future will need to think hard about how to balance remote and on-site readings for the benefit of patients, education, and cross-fertilization of knowledge. It will be important for radiologists to maintain a strong presence within their health care institutions so that they continue to be recognized as key partners in patient care and research.Artificial IntelligenceAI will not replace radiology, but it will profoundly affect our relevance and our workflow. The greatest risk to our specialty is that other medical specialties, and potentially patients themselves, may leverage AI for independent image interpretation. Given that referring physicians have clinical information at their fingertips, they will be able to put imaging findings in a clinically relevant context to a much greater degree than traditional radiologists.The globalization of consultative work combined with AI automation will increasingly threaten many white-collar disciplines. These include medicine and law (5). Our pathology colleagues have seen evolution in practice. For example, the Papanicolaou test no longer requires human interpretation routinely (6). Machine-based interpretations, potentially in distant locations, have become the norm. The only exceptions are in unusual conditions that prevent automatic reading and require human intervention. Certain radiology examinations will undergo a similar evolution. AI algorithms will provide comprehensive and autonomous interpretation. At present, most AI algorithms are narrowly focused, targeting a specific imaging feature or function. Over the next several years, algorithm development will broaden substantially to include a comprehensive evaluation of all possible features that may be present in certain imaging examinations.Algorithm development will become increasingly federated, with the training of distributed algorithms on local data rather than moving data beyond secure firewalls. This will increase the diversity of training data and minimize training bias, thereby promoting health equity and algorithm robustness. Once an algorithm is trained, its quality will be maintained with continuous learning. Temporal degradation owing to changes in local data environments will be minimal (7,8). Radiologists must remain the keepers of such algorithms and oversee their use unequivocally.Just as AI will permeate all facets of daily living, it will affect all aspects of radiology practice. These will range from examination triage and planning to lesion detection, characterization, and measurement (9). AI will also enable greater assessment of disease likelihood and potential treatment outcomes by seamless integration of imaging findings with other clinical indicators. Radiologists’ role in this continuum is not assured. We must see to it that we are adding value at every step.The “lifelong learning” of AI algorithms, including their ability to adapt to improved or new technologies (eg, novel MRI pulse sequences, novel radiotracers) will influence the training of the next generation of radiologists. The integration of AI into training will continuously evolve. Radiologists will, at least in the short- and mid-term, still be required to supervise and, if necessary, overrule AI, thereby providing feedback and improving model performance. While training supported by AI will surely be a great learning opportunity, especially when visualization tools such as gradient-weighted class activation mapping, or Grad-CAM, are applied, we need to ensure that trainees acquire sufficient knowledge and do not unquestioningly trust the results produced by an algorithm (10). For this, trainees need at least a basic understanding of AI algorithm design and training strategies to understand AI algorithms’ performance levels and limitations.Over time, AI algorithms will enable the integration of radiomic features and metabolic or functional information with genomic and other phenotypic information for disease detection and characterization. These capabilities will likely exceed what is achievable by humans alone (11). As radiologists, we must welcome the assistance that AI will provide to our practices, recognizing that it will enable us to function at a higher level, just as did the advent of digital imaging and electronic image display. To truly harness this AI-enhanced, integrated diagnostic approach, cross-training with other specialties such as pathology will be vital for the next generation of radiology trainees. This may be achieved by electives during residency, short-term fellowships, or at an even earlier point during medical school in the form of integrated MD-PhD programs.Value-based CareNon–image-based precision diagnostics (ie, “liquid biopsies”) will continue to rise, and radiology’s place in the value equation will need to evolve commensurately. We must strive to be the primary purveyors of imaging examinations in every respect, from their most appropriate uses to timely, accurate, and precise reporting of the information they contain. Imaging centers will evolve beyond diagnostic centers to treatment planning and prediction centers. Radiology will be responsible for phenotyping with imaging markers that are prognostic and predictive of treatment response on a large scale.The drive toward subspecialization for maximal value creation will continue. But AI may enable radiologists to achieve highly subspecialized excellence (as so-called centaur radiologists—a combination of human plus computer) (12). While it is tempting to imagine that this scenario may enable AI-powered generalists to deliver the value typically seen with subspecialists, subspecialized training will remain extremely important, as radiologists’ clinical relevance can only be assured by radiologists being as knowledgeable as our subspecialized referring physicians.Imaging examinations will give rise to AI-derived three-dimensional data sets of imaging findings. Radiologists will need to ensure the accuracy of salient findings. More importantly, radiologists must ensure the integration of these findings with relevant clinical scenarios to produce meaningful and impactful diagnoses. Imaging interpretations will be converted to lay language instantly and in a variety of media, including static and motion content with automatic, cinematically rendered images (13). Patients and their providers will have instant and full access to their images and interpretations in real time, potentially for the application of individual or personal AI algorithms on features of interest. Recommended follow-up imaging or other diagnostic testing will be communicated, confirmed, and, if warranted, arranged automatically through robust care coordination systems (14). These changes will further marginalize radiologists and threaten our relevance unless we integrate ourselves in the care continuum and add value beyond machine-generated interpretations and recommendations.Environmental, Social, and Governance ConcernsThe carbon footprint of medical imaging will be at the forefront of environmental sustainability efforts. Thus, technical advances that substantially reduce our energy consumption will be embraced uniformly. These include imaging devices and viewing equipment with reduced power requirements. Renewable sources for consumables in diagnostic and interventional radiology will become mainstream.Radiology’s role in ensuring that health care is delivered equitably across all demographic groups of our society is not assured. Economic pressures will continue to favor the greatest care delivery to those who can most afford it. We must remain vigilant to do our part in bringing the power of medical imaging to the most underserved populations across the globe (9).Similarly, diversity among our leadership and within our governance structures is not assured. We must continually strive for parity for women and members of other underrepresented groups, or the axiom “if you’re not at the table, you’re on the menu” will be as apt in 2040 as it is today.Evolution of Imaging TechnologiesImaging technologies will become increasingly multiscale, multimodal, and multiomic. They will incorporate relevant metabolic, proteomic, and genomic information. “Smart” contrast agents will increase exponentially, leveraging in vivo biochemistry, immunohistochemistry, reporter genes (genes that produce receptors that bind imaging probes), and nanoparticles. Bioengineering will advance rapidly, showing explosive potential to alter the practice of medicine, with an outsized impact on medical imaging. Phenotypic data generated by the physical interaction of an external energy source with biologic tissue will enter multiomic databases that cross multiple levels of biologic regulation, potentially bypassing the need for image formation (ie, fingerprinting). Images will be reconstructed in select cases where human understanding is best achieved by visual means, such as for surgical or interventional radiology treatment planning.Therapy: An Expanding Role of RadiologyPrecision Imaging and Image-guided InterventionDue to advances in imaging, devices, techniques, and robotics, the array of minimally invasive interventions will continue to grow (1). Percutaneous endoscopic imaging will enable interventional radiologists to perform minimally invasive interventions in domains previously reserved for other specialties. When combined with multispectral optical imaging, in vivo virtual histologic examination may be possible for certain pathologic conditions, potentially obviating the need for a needle biopsy (15).Interventional procedures will become increasingly automated, powered by AI. This will enable less-trained personnel to carry out some less complicated image-guided procedures, potentially at home and other locations beyond the purview of standard health care facilities (16). AI-powered robotics may follow for more complex procedures.RadiotheranosticsMolecular imaging will continue to expand through basic science research and clinical practice via the rapid rise in theranostics. Theranostics combines diagnostic imaging with targeted therapy to noninvasively determine tumor phenotype and evaluate functional and molecular responses to therapy.Radiotheranostics will continue to expand exponentially, combining molecular imaging—currently with PET and SPECT—with radionuclide therapy, leveraging small drugs, peptides, and antibodies to carry therapeutic radionuclides (alpha-, beta-, or auger-emitters). Radiotheranostic growth will expand clinicians’ ability to first “see with precision” and then “treat with targeting.” While oncologic radiotheranostics already includes numerous contemporary applications, it has enormous untapped potential for treating a huge range of cancers, given its capacity to be adapted to different targets and maximize tumor killing while minimizing side effects (17).Real-time feedback from imaging allows clinicians to improve patient selection for therapy, provide earlier and more precise treatment response assessment, and expedite drug development. Whereas patients are typically selected for cancer therapies based on clinical (and secondarily histopathologic or molecular) parameters, pairing radiation therapy with a companion diagnostic (eg, technetium 99m, fluorine 18, or gallium 68) allows clinicians to visualize in vivo the expression of the therapeutic target, confirm the presence of the therapeutic agent, and visualize tumor burden. In the field of oncology, drug development currently fails about 90% of the time. Radionuclide therapies, however, provide insights regarding the biodistribution of both intended targets and radioligands that allow researchers to quickly halt, adapt, or accelerate their experiments. This increases the success rates of these therapies over those of conventional cancer therapies.The development and implementation of radiotheranostics will clearly represent a tremendous opportunity for radiology and nuclear medicine to take leading roles in cancer diagnosis and treatment. Furthermore, while radionuclide probes for the selective and targeted diagnosis and treatment of specific cells hold immense promise for cancer, they also will have potential for application to other patient-specific diseases, including inflammatory and autoimmune conditions. Realizing the full potential of radiotheranostics for advancing clinical care will require more extensive cross-training of radiologists, and likely other clinicians, in nuclear medicine (18).ConclusionThe next 2 decades will bring astonishing changes to health care that are yet to be imagined. With the expected growth of AI, radiologists will need to offer added value beyond image interpretation. We have been tremendously innovative over time, and we must remain at the leading edge of innovation in both diagnostic imaging and image-guided therapy (eg, interventional radiology and radiotheranostics) to ensure that our technical advances and expertise continue to provide value in the years ahead. We are confident that with flexibility and agility, we will sustain our relevance to the full continuum of clinical care. However, we must continue to embrace changes that are best for patients, and decisions made with patients at their center will always serve us well. Radiologists will carry on shaping our own destiny just as we have for the past 100 years, since the first innovations of our specialty were set to print in Radiology.Disclosures of conflicts of interest: J.A.B. Member of the board of directors for Accumen and president of the International Society for Strategic Studies in Radiology. H.H. Grants from the National Cancer Institute (principal investigator or co-leader); member of the external advisory board of the Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins, the international advisory board of the University of Vienna, the Scientific Committee of the DKFZ (German Cancer Research Center), the board of trustees of the DKFZ, and the board of directors of Paige; paid member of the board of directors of Ion Beam Applications.AcknowledgmentsThe authors thank Jason Lewis, PhD; Marius Mayerhoefer, MD, PhD; and Ralph Weissleder, MD, PhD, for their valuable input and Ada Muellner, MS, for editing the manuscript.References1. Weissleder R. Leveraging technology to address current and future world issues. Sci Transl Med 2022;14(666):eabq0606. Crossref, Medline, Google Scholar2. Akkus Z, Cai J, Boonrod A, et al. A survey of deep-learning applications in ultrasound: artificial intelligence-powered ultrasound for improving clinical workflow. J Am Coll Radiol 2019;16(9 Pt B):1318–1328. Crossref, Medline, Google Scholar3. Cooley CZ, Stockmann JP, Armstrong BD, et al. Two-dimensional imaging in a lightweight portable MRI scanner without gradient coils. Magn Reson Med 2015;73(2):872–883. Crossref, Medline, Google Scholar4. Cramer A, Hecla J, Wu D, et al. Stationary computed tomography for space and other resource-constrained environments. Sci Rep 2018;8(1):14195. Crossref, Medline, Google Scholar5. Baldwin R. The Globotics Upheaval: Globalization, Robotics, and the Future of Work. New York, NY: Oxford University Press, 2019. Google Scholar6. Karasu Benyes Y, Welch EC, Singhal A, Ou J, Tripathi A. A comparative analysis of deep learning models for automated cross-preparation diagnosis of multi-cell liquid Pap smear images. Diagnostics (Basel) 2022;12(8):1838. Crossref, Medline, Google Scholar7. Hosny A, Parmar C, Quackenbush J, Schwartz LH, Aerts HJWL. Artificial intelligence in radiology. Nat Rev Cancer 2018;18(8):500–510. Crossref, Medline, Google Scholar8. Pianykh OS, Langs G, Dewey M, et al. Continuous learning AI in radiology: implementation principles and early applications. Radiology 2020;297(1):6–14. Link, Google Scholar9. Hricak H, Abdel-Wahab M, Atun R, et al. Medical imaging and nuclear medicine: a Lancet Oncology Commission. Lancet Oncol 2021;22(4):e136–e172. Crossref, Medline, Google Scholar10. Oya M, Sugimoto S, Sasai K, Yokoyama K. Investigation of clinical target volume segmentation for whole breast irradiation using three-dimensional convolutional neural networks with gradient-weighted class activation mapping. Radiol Phys Technol 2021;14(3):238–247. Crossref, Medline, Google Scholar11. Thrall JH, Li X, Li Q, et al. Artificial intelligence and machine learning in radiology: opportunities, challenges, pitfalls, and criteria for success. J Am Coll Radiol 2018;15(3 Pt B):504–508. Crossref, Medline, Google Scholar12. Dreyer KJ, Geis JR. When machines think: radiology’s next frontier. Radiology 2017;285(3):713–718. Link, Google Scholar13. Recht MP, Westerhoff M, Doshi AM, et al. Video radiology reports: a valuable tool to improve patient-centered radiology. AJR Am J Roentgenol 2022;219(3):509–519. Crossref, Medline, Google Scholar14. Lacson R, Gujrathi I, Healey M, et al. Closing the loop on unscheduled diagnostic imaging orders: a systems-based approach. J Am Coll Radiol 2021;18(1 Pt A):60–67. Crossref, Medline, Google Scholar15. Weissleder R, Nahrendorf M. Advancing biomedical imaging. Proc Natl Acad Sci U S A 2015;112(47):14424–14428. Crossref, Medline, Google Scholar16. Brattain LJ, Pierce TT, Gjesteby LA, et al. AI-enabled, ultrasound-guided handheld robotic device for femoral vascular access. Biosensors (Basel) 2021;11(12):522. Crossref, Medline, Google Scholar17. Bodei L, Herrmann K, Schöder H, Scott AM, Lewis JS. Radiotheranostics in oncology: current challenges and emerging opportunities. Nat Rev Clin Oncol 2022;19(8):534–550. Crossref, Medline, Google Scholar18. Hricak H. 2016 New Horizons Lecture: beyond imaging—radiology of tomorrow. Radiology 2018;286(3):764–775. Link, Google ScholarArticle HistoryReceived: Oct 10 2022Revision requested: Oct 14 2022Revision received: Oct 20 2022Accepted: Oct 25 2022Published online: Dec 19 2022Published in print: Jan 2023 FiguresReferencesRelatedDetailsCited ByChange Is Good: The Evolution and Future of Breast ImagingLinda Moy, 7 February 2023 | Radiology, Vol. 0, No. 0Accompanying This ArticleRadiology 2040 - Perspective from Dr. Hedvig HricakJan 31 2023Default Digital Object SeriesRadiology 2040 - Perspective from Dr. James BrinkFeb 7 2023Default Digital Object SeriesRecommended Articles RSNA Education Exhibits RSNA Case Collection Vol. 306, No. 1 PodcastPodcastSupplemental MaterialMetrics Altmetric Score PDF download
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
更新
大幅提高文件上传限制,最高150M (2024-4-1)

科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
科研通AI2S应助南滨采纳,获得10
2秒前
SunnyYim完成签到,获得积分10
2秒前
轻松小之发布了新的文献求助30
2秒前
CC完成签到,获得积分10
3秒前
CuSO4完成签到,获得积分10
3秒前
科研通AI2S应助shi hui采纳,获得10
3秒前
water完成签到 ,获得积分10
3秒前
PG完成签到 ,获得积分10
5秒前
南北完成签到 ,获得积分10
6秒前
古的古的应助榕树采纳,获得10
6秒前
曼曼来发布了新的文献求助10
6秒前
懵了应助周周采纳,获得30
8秒前
9秒前
03210322完成签到 ,获得积分10
11秒前
zickycc完成签到,获得积分20
12秒前
Raymond完成签到,获得积分10
13秒前
james完成签到,获得积分10
13秒前
14秒前
XLH发布了新的文献求助10
15秒前
15秒前
LL完成签到,获得积分10
15秒前
rain完成签到,获得积分10
16秒前
shuangyanli完成签到,获得积分10
17秒前
ice贝完成签到,获得积分10
18秒前
周周完成签到,获得积分10
19秒前
啵啵完成签到,获得积分10
20秒前
王aa关注了科研通微信公众号
21秒前
大红豆子完成签到,获得积分10
21秒前
23秒前
子车茗应助Nancy采纳,获得30
25秒前
SciKid524完成签到 ,获得积分10
27秒前
zhangxiaohua完成签到,获得积分10
29秒前
小西瓜发布了新的文献求助10
29秒前
端庄栾发布了新的文献求助10
29秒前
quzhenzxxx完成签到 ,获得积分10
29秒前
852应助服部平次采纳,获得10
31秒前
MLi应助天想月采纳,获得10
33秒前
连牙蓝上了吗完成签到 ,获得积分10
33秒前
35秒前
草莓糖发布了新的文献求助10
35秒前
高分求助中
Production Logging: Theoretical and Interpretive Elements 2000
Very-high-order BVD Schemes Using β-variable THINC Method 1200
RNAの科学 ―時代を拓く生体分子― 金井 昭夫(編) 1000
BIOLOGY OF NON-CHORDATES 1000
进口的时尚——14世纪东方丝绸与意大利艺术 Imported Fashion:Oriental Silks and Italian Arts in the 14th Century 800
Autoregulatory progressive resistance exercise: linear versus a velocity-based flexible model 550
Education and Upward Social Mobility in China: Imagining Positive Sociology with Bourdieu 500
热门求助领域 (近24小时)
化学 医学 生物 材料科学 工程类 有机化学 生物化学 物理 内科学 纳米技术 计算机科学 化学工程 复合材料 基因 遗传学 物理化学 催化作用 细胞生物学 免疫学 冶金
热门帖子
关注 科研通微信公众号,转发送积分 3353809
求助须知:如何正确求助?哪些是违规求助? 2978311
关于积分的说明 8685219
捐赠科研通 2659860
什么是DOI,文献DOI怎么找? 1456376
科研通“疑难数据库(出版商)”最低求助积分说明 674342
邀请新用户注册赠送积分活动 665114