AI boom in medical imaging … a data perspective

Medical Imaging data is growing rapidly from ~400PetaBytes in 2020 to more than 630PetaBytes in 2030?

Imagine a world where every second, countless images of the human body are captured and stored, revealing hidden ailments and saving lives. This isn't science fiction—it's the reality of modern medical imaging. Over the past decade, the amount of data generated by X-rays, CT scans, MRIs, and ultrasounds has skyrocketed, creating a data explosion that's transforming healthcare. From the bustling hospitals in big-city America to remote clinics in rural India, medical imaging data is growing at an unprecedented rate, driving innovations in diagnosis and treatment that were once unimaginable.?

Welcome to the era of exponential growth in medical imaging, where data is not just information, but the key to unlocking a healthier future for all with artificial intelligence/machine learning (AI-ML)? enabled devices.?

AI ubiquity has been in the news a lot lately. The FDA recently released its list of cleared medical devices with AI. With 76% (671 out of 882) of? FDA cleared AI products, Radiology is clearly leading the charge in AI in healthcare. While many of these AI applications are focused on historical discriminative AI methods, a new generation of products with service-as-a-software paradigms are being developed based on the creativity, context and reasoning abilities of foundation models for radiology. These models are helping with earlier notification of patient mal conditions, better care coordination and improved administrative workflows in health systems and clinics. As medical imaging specific foundation models get built, it would be instructive to understand the growth of medical imaging data by modality (ultrasound, MR, CT, X-ray), by geography over the past decade and project this growth into the coming decade. This project is a collaboration with GPT-4. As a disclaimer, the data sources need further analyses. I am putting this article out to get feedback from the highly informed LinkedIn network.?

Here's the approach to the analyses.

1. Estimate the data size per imaging modality
2. Determine how often the imaging procedures are performed
3. Estimate the geographical distribution and growth rate of imaging procedures
4. Baseline the number of medical images based on available 2010 data
5. Project the medical imaging data growth to 2030

1. Data sources

Medical Imaging Statistics: The International Atomic Energy Agency (IAEA) and World Health Organization (WHO) provide statistics on the number of imaging procedures performed globally. Reports from the Radiological Society of North America (RSNA) and the American College of Radiology (ACR) offer insights into imaging procedure volumes and trends. Market research reports from sources such as Frost & Sullivan and MarketsandMarkets provide data on the growth rates of different imaging modalities.

Average File Sizes for Imaging Modalities: Peer-reviewed articles and technical specifications from medical imaging equipment manufacturers, such as Siemens, GE Healthcare, and Philips, provide data on average file sizes for different modalities. Clinical radiology textbooks and publications also give detailed information on typical data volumes for various imaging procedures.

Growth Rates and Projections: Historical growth rates for medical imaging can be found in health care market analysis reports from organizations like the International Data Corporation (IDC) and reports from healthcare consultancies such as Deloitte and McKinsey.Government health departments and international health organizations often publish annual reports that track changes in healthcare services, including imaging.

Regional Distribution: The OECD Health Statistics database provides detailed health data for member countries, including the number of imaging procedures.National health services, such as the Centers for Disease Control and Prevention (CDC) in the United States and the National Health Service (NHS) in the UK, publish annual reports on healthcare services utilization, including imaging.

Projected Trends and Technological Advances: Journals such as the Journal of Digital Imaging and the American Journal of Roentgenology (AJR) publish research on trends and advancements in medical imaging technology.

Industry white papers from leading medical imaging companies discuss future trends and projected increases in imaging procedures. Specific references are listed in the bibliography section.

2. Data Size per Imaging Modality

First, we estimate the average size of the data generated by each type of medical imaging modality:

• X-ray: An X-ray image typically ranges from 10 MB to 20 MB.
• Ultrasound: Ultrasound image data sizes depend on the type of studies. Echo exams of the heart can include loops/videos up to 1GB. Obstetrics images can range to 100-300 MB. However, a majority of radiology images tend to capture still images averaging around 3 MB to 30 MB per study.
• Computed Tomography (CT): A single CT scan can generate between 200 MB and 1 GB, depending on the number of slices and resolution.
• Magnetic Resonance Imaging (MRI): An MRI scan can produce between 100 MB and 500 MB, depending on the sequences used and the resolution.

3. Frequency of Imaging Procedures

The four modalities that dominate the medical imaging procedures include X-ray, Ultrasound, CT and MR. For these analyses, we will ignore the rest of the imaging modalities like nuclear medicine imaging (PET/NMR), Mammography, CathLab and other surgical visualization procedures.?

• X-rays: Approximately 3.6 billion X-ray examinations are performed globally each year.
• Ultrasound: There are approximately 1.5 billion ultrasound examinations done each year.
• CT scans: There are around 450 million CT scans conducted annually worldwide.
• MRI scans: About 150 million MRI scans are performed each year globally.

4. Geographic Distribution

Healthcare infrastructure and imaging practices vary across different regions. We estimate the following:

• United States: High usage of advanced imaging modalities (e.g., CT, MRI). The U.S. accounts for about 20% of global medical imaging.
• China: Rapidly growing medical imaging sector, contributing around 15% of global imaging.
• Asia-Pacific (excluding China): Significant growth, especially in countries like Japan, India, and South Korea, contributing about 30% of global imaging.
• Europe: Well-established medical imaging infrastructure, accounting for about 25% of global imaging.
• Rest of the World: Includes regions with varied healthcare systems, contributing around 10% of global imaging.

5. Baseline data (2010)

We'll use approximate data from 2010 as the base to calculate growth:

• X-rays (2010): 2.7 billion exams
• Ultrasound (2010): 1 billion exams
• CT scans (2010): 280 million exams
• MRI scans (2010): 85 million exams

6. Historical Growth Rates and Growth calculations

We can use available data and reports to estimate the annual growth rates for medical imaging procedures globally. Historically, the medical imaging market has grown at different rates depending on the modality and region. We will estimate ~3% for X-ray imaging, ~4% for ultrasound, ~5% for CT scans and ~6% for MR imaging. Using the base data and growth rates, we calculate the number of exams and data generated for each subsequent year until 2030 (rounded for simplicity):

X-rays:

• 2010: 2.7 billion exams × 15 MB = 40.5 PB
• 2020: 2.7 billion × (1.03^10) = 3.62 billion exams × 15 MB = 54.3 PB
• 2030: 2.7 billion × (1.03^20) = 4.86 billion exams × 15 MB = 72.9 PB

Ultrasound:

• 2010: 1 billion exams × 15 MB = 15 PB
• 2020: 1 billion × (1.04^10) = 1.48 billion exams × 15 MB = 22.2 PB
• 2030: 1 billion × (1.04^20) = 2.19 billion exams × 15 MB = 32.85 PB

CT scans:

• 2010: 280 million exams × 600 MB = 168 PB
• 2020: 280 million × (1.05^10) = 456 million exams × 600 MB = 273.6 PB
• 2030: 280 million × (1.05^20) = 743 million exams × 600 MB = 445.8 PB

MRI scans:

• 2010: 85 million exams × 300 MB = 25.5 PB
• 2020: 85 million × (1.06^10) = 152 million exams × 300 MB = 45.6 PB
• 2030: 85 million × (1.06^20) = 272 million exams × 300 MB = 81.6 PB

7. Summing Up Yearly Data

We'll sum the data generated for each modality for the key years (2010, 2020, 2030):

2010 Total:

• X-rays: 40.5 PB
• CT: 168 PB
• MRI: 25.5 PB
• Ultrasound: 15 PB
• Total: 249 PB

2020 Total:

• X-rays: 54.3 PB
• CT: 273.6 PB
• MRI: 45.6 PB
• Ultrasound: 22.2 PB
• Total: 395.7 PB

2030 Total:

• X-rays: 72.9 PB
• CT: 445.8 PB
• MRI: 81.6 PB
• Ultrasound: 32.85 PB
• Total: 633.15 PB

8. Calculating Total Data Volume

Let's calculate the total data volume for each modality and region using 2020 procedure volume as a baseline (rounding down):

X-rays:

• Global: 3.6 billion exams × 15 MB (average) = 54 petabytes (PB)
• U.S.: 0.72 billion exams × 15 MB = 10.8 PB
• China: 0.54 billion exams × 15 MB = 8.1 PB
• APAC: 1.08 billion exams × 15 MB = 16.2 PB
• Europe: 0.9 billion exams × 15 MB = 13.5 PB
• Rest of the World: 0.36 billion exams × 15 MB = 5.4 PB

Ultrasound:

• Global: 1.5 billion exams × 15 MB (average) = 22.5 PB
• U.S.: 0.3 billion exams × 15 MB = 4.5 PB
• China: 0.225 billion exams × 15 MB = 3.375 PB
• APAC: 0.45 billion exams × 15 MB = 6.75 PB
• Europe: 0.375 billion exams × 15 MB = 5.625 PB
• Rest of the World: 0.15 billion exams × 15 MB = 2.25 PB

CT scans:

• Global: 450 million exams × 600 MB (average) = 270 PB
• U.S.: 90 million exams × 600 MB = 54 PB
• China: 67.5 million exams × 600 MB = 40.5 PB
• APAC: 135 million exams × 600 MB = 81 PB
• Europe: 112.5 million exams × 600 MB = 67.5 PB
• Rest of the World: 45 million exams × 600 MB = 27 PB

MRI scans:

• Global: 150 million exams × 300 MB (average) = 45 PB
• U.S.: 30 million exams × 300 MB = 9 PB
• China: 22.5 million exams × 300 MB = 6.75 PB
• APAC: 45 million exams × 300 MB = 13.5 PB
• Europe: 37.5 million exams × 300 MB = 11.25 PB
• Rest of the World: 15 million exams × 300 MB = 4.5 PB

9. Summing Up

Now, let's aggregate the data:

Total Global Data:

• X-rays: 54 PB
• Ultrasound: 22.5 PB
• CT: 270 PB
• MRI: 45 PB
• Total: 391.5 PB

Regional Breakdown (approximate sums):

• United States: 10.8 PB (X-ray) + 54 PB (CT) + 9 PB (MRI) + 4.5 PB (Ultrasound) = 78.3 PB
• China: 8.1 PB (X-ray) + 40.5 PB (CT) + 6.75 PB (MRI) + 3.375 PB (Ultrasound) = 58.725 PB
• APAC: 16.2 PB (X-ray) + 81 PB (CT) + 13.5 PB (MRI) + 6.75 PB (Ultrasound) = 117.45 PB
• Europe: 13.5 PB (X-ray) + 67.5 PB (CT) + 11.25 PB (MRI) + 5.625 PB (Ultrasound) = 97.875 PB
• Rest of the World: 5.4 PB (X-ray) + 27 PB (CT) + 4.5 PB (MRI) + 2.25 PB (Ultrasound) = 39.15 PB

Conclusion

Globally, ~390 petabytes of data are generated annually from medical imaging. The United States and China are major contributors, with the Asia-Pacific region showing significant growth in data generation. This estimate highlights the substantial and growing volume of data produced in the medical imaging field, underscoring the importance of efficient data management and storage solutions in healthcare.

The data generated by medical imaging modalities globally has increased substantially from 249 petabytes in 2010 to an estimated 390 petabytes in 2020. By 2030, it is projected to reach approximately 633 petabytes. This projection is based on historical growth rates and assumes that these rates will remain relatively stable, factoring in advancements in imaging technology and increased accessibility to medical imaging services globally.

To provide a more comprehensive year-by-year breakdown or detailed growth for each region, further data specific to annual increases and regional variances would be necessary, but the above methodology provides a robust estimate based on available data and reasonable growth assumptions.

Of course, these data may be locked up in disjointed databases in disparate health-systems and clinics and not easily available to technology companies for foundation model development and refinement. At the recent Stanford RAISE health symposium, many medical informatics experts lamented that scaling the potential of AI/ML solutions in their health systems, still requires much work in moving from billing-based medical procedures to truly outcomes oriented medical practice.

Bibliography:

1. US FDA (May 13, 2024 update). Artificial intelligence and machine learning (AI/ML)-Enabled medical devices. https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-aiml-enabled-medical-devices
2. Patient radiation exposure monitoring in medical imaging. 2023 https://www.iaea.org/publications/14971/patient-radiation-exposure-monitoring-in-medical-imaging
3. Hricak H et al Lancet oncology commision on medical imaging and nuclear medicine Lancet Oncol. 2021 Apr; 22(4): e136–e172.
4. Commission CQ. Radiology review: a national review of radiology reporting within the NHS in England. Care Quality Commission. 2018. https://www.cqc.org.uk/sites/default/files/20180718-radiology-reporting-review-report-final-for-web.pdf
5. DeStigter K, Pool K-L, Leslie A, Hussain S, Tan BS, Donoso-Bach L, et al. Optimizing integrated imaging service delivery by tier in low-resource health systems. Insights into Imaging. 2021;12(1):1–11. https://insightsimaging.springeropen.com/articles/10.1186/s13244-021-01073-8
6. Lau L. Leadership and management in quality radiology. Biomed Imaging Interv J. 2007;3(3): 1-7. https://pubmed.ncbi.nlm.nih.gov/21614284/
7. WHO. Delivering quality health services: a global imperative. OECD Publishing; 2018. https://apps.who.int/iris/handle/10665/272465
8. WHO. Global atlas of medical devices. 2017. https://apps.who.int/iris/handle/10665/255181
9. Hong AS et al Trends in diagnostic imaging utilization among medicare and commercially insured adults from 2003 through 2016 (2019). https://pubs.rsna.org/doi/full/10.1148/radiol.2019191116
10. Brady AP et al Radiology in the era of Value-based Healthcare: A multi-society expert statement from the ACR, CAR, ESR, IS3R, RANZCR, and RSNA https://pubs.rsna.org/doi/full/10.1148/radiol.2020209027
11. McEnery KW et al. Predictive modeling for future CT imaging procedure volume - recovering from Harvey https://www.acr.org/-/media/ACR/NOINDEX/Abstracts/2018/18031_McEnery.pdf
12. ?Davenport MS et al CT volume from 2398 Radiology practices in the United States: A real-time indicator of the effect of COVID-19 on Routine care, January to September 2020. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7577702/
13. Maximum capacity: Overloaded radiologist are grappling with solutions to a booming volume crisis (2024) https://www.acr.org/Practice-Management-Quality-Informatics/ACR-Bulletin/Articles/April-2024/Maximum-Capacity
14. OECD Health Statistics: MRI units: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7577702/
15. OECD Health Statistics: CT exams: https://data.oecd.org/healthcare/computed-tomography-ct-exams.htm
16. MarketsandMarkets. "Diagnostic imaging market worth $34.6 billio by 2028" https://www.marketsandmarkets.com/PressReleases/diagnostic-imaging-market.asp
17. Diagnostic imaging market by product (MRI (Open, Closed)), Ultrasound (2D, 4D, Doppler), X-Ray (Digital, Analog), CT, SPECT, Hybrid PET, Mammography). Application (OB/GYN, CVDs, Brain, Cancer), End user (Hospitals, Clinics), and Region - Global Forecast to 2028 https://www.marketsandmarkets.com/Market-Reports/diagnostic-imaging-market-411.html
18. Frost & Sullivan. "Global Medical Imaging and informatics outlook, 2022" https://store.frost.com/global-medical-imaging-and-informatics-outlook-2022.html
19. World industry news: Diagnostic imaging market size, share, trends and projected growth https://www.dhirubhai.net/pulse/diagnostic-imaging-market-size-share-trends-projected-hgfmf/