This statistic depicts the annual compensation among radiologists in the U.S. according to different sources (organizations), as of 2018. According to Integrated Healthcare Strategies, annual salaries for radiologists averaged some *** thousand U.S. dollars.

As of April 2025, there were a total of 570,655 specialty physicians active in the United States. Of these, most were specialized in emergency medicine. Physician compensation Significant pay variations exist across specialties and regions, with orthopedic doctors and surgeons command the highest average annual salaries at 564,000 U.S. dollars. Meanwhile, the Midwest region offers the highest average physician compensation at 385,000 U.S. dollars annually. Interestingly, doctors in Northern parts of the United States tend to earn less than their counterparts in other regions. Burnout among physicians Despite high salaries, U.S. physicians face high workload and stress in the workplace. Nearly half of surveyed doctors reported feeling burnout, with higher burnout rates among female doctors, younger physicians, and those in primary care compared to their counterparts. More effort to combat burnout is needed in the healthcare system. Increasing compensation was cited by physicians as the top measure to alleviate burnout, followed by adding support staff and offering more flexible schedules.

The MIMIC Chest X-ray JPG (MIMIC-CXR-JPG) Database v2.0.0 is a large publicly available dataset of chest radiographs in JPG format with structured labels derived from free-text radiology reports. The MIMIC-CXR-JPG dataset is wholly derived from MIMIC-CXR, providing JPG format files derived from the DICOM images and structured labels derived from the free-text reports. The aim of MIMIC-CXR-JPG is to provide a convenient processed version of MIMIC-CXR, as well as to provide a standard reference for data splits and image labels. The dataset contains 377,110 JPG format images and structured labels derived from the 227,827 free-text radiology reports associated with these images. The dataset is de-identified to satisfy the US Health Insurance Portability and Accountability Act of 1996 (HIPAA) Safe Harbor requirements. Protected health information (PHI) has been removed. The dataset is intended to support a wide body of research in medicine including image understanding, natural language processing, and decision support.

"Facilitate marketing campaigns with the healthcare email list from Infotanks Media that includes doctors, healthcare professionals, NPI numbers, physician specialties, and more. Buy targeted email lists of healthcare professionals and connect with doctors, specialists, and other healthcare professionals to promote your products and services. Hyper personalize campaigns to increase engagement for better chances of conversion. Reach out to our data experts today! Access 1.2 million physician contact database with 150+ specialities including chiropractors, cardiologists, psychiatrists, and radiologists among others. Get ready to integrate healthcare email lists from Infotanks Media to start email marketing campaigns through any CRM and ESP. Contact us right now! Ensure guaranteed lead generation with segmented email marketing strategies for specialists, departments, and more. Make the best use of target marketing to progress and move closer to your business goals with email listing services for healthcare professionals. Infotanks Media provides 100% verified healthcare email lists with the highest email deliverability guarantee of 95%. Get a custom quote today as per your requirements. Enhance your marketing campaigns with healthcare email lists from 170+ countries to build your global outreach. Request your free sample today! Personalize your business communication and interactions to maximize conversion rates with high quality contact data. Grow your business network in your target markets from anywhere in the world with a guaranteed 95% contact accuracy of the healthcare email lists from Infotanks Media. Contact data experts at Infotanks Media from the healthcare industry to get a quick sample for free. Write to us or call today!

Hyper target within and outside your desired markets with GDPR and CAN-SPAM compliant healthcare email lists that get integrated into your CRM and ESPs. Balance out the sales and marketing efforts by aligning goals using email lists from the healthcare industry. Build strong business relationships with potential clients through personalized campaigns. Call Infotanks Media for a free consultation. Explore new geographies and target markets with a focused approach using healthcare email lists. Align your sales teams and marketing teams through personalized email marketing campaigns to ensure they accomplish business goals together. Add value and grow revenue to take your business to the next level of success. Double up your business and revenue growth with email lists of healthcare professionals. Send segmented campaigns to monitor behaviors and understand the purchasing habits of your potential clients. Send follow up nurturing email marketing campaigns to attract your potential clients to become converted customers. Close deals sooner with detailed information of your prospects using the healthcare email list from Infotanks Media. Reach healthcare professionals on their preferred platform of communication with the email list of healthcare professionals. Identify, capture, explore, and grow in your target markets anywhere in the world with a fully verified, validated, and compliant email database of healthcare professionals. Move beyond the traditional approach and automate sales cycles with buying triggers sent through email marketing campaigns. Use the healthcare email list from Infotanks Media to engage with your targeted potential clients and get them to respond. Increase email marketing campaign response rate to convert better! Reach out to Infotanks Media to customize your healthcare email lists. Call today!"

Abstract

This dataset includes MR imaging from 203 glioma patients with 617 different post-treatment MR time points, and tumor segmentations. Clinical data includes patient demographics, genomics, and treatment details. Preprocessing of MR images followed a standardized pipeline with automatic tumor segmentation based on nnUNet deep learning approach. The automatic tumor segmentations were manually validated and refined by neuroradiologists.

The heterogeneity of glioma imaging characteristics and management strategies contributes to a lack of reliable findings when evaluating treatment outcomes with conventional MRI, and the overlapping imaging features of radiation necrosis and tumor progression post-treatment can be particularly challenging for radiologists. This robust dataset should contribute to the development of AI models to improve evaluation of treatment outcomes.

Introduction

The dataset consists of institutional review board-approved retrospective analysis of pathologically proven glioma patients at University Hospital of The University of Missouri - Anatomic Pathology CoPathPlus database was used to collect glioma cases over the last 10 years.

Sharing segmented postoperative glioma data with clinical information significantly accelerates research and improves clinical practice by providing a comprehensive, readily available dataset. This eliminates the time-consuming burden of manual segmentation, enhances the accuracy and consistency of tumor delineation, and allows researchers to focus on analysis and interpretation, ultimately driving the development of more accurate segmentation algorithms, predictive models for personalized treatment strategies, and improved patient outcome predictions. Standardized longitudinal follow-up and benchmarking capabilities further facilitate multi-center studies and objective evaluation of treatment efficacy, leading to advancements in glioma biology and personalized patient care.

Methods

The following subsections provide information about how the data were selected, acquired, and prepared for publication.

Subject Inclusion and Exclusion Criteria

The selection criteria for the CoPath Natural Language II Search included accession dates ranging from 01/01/2021 to 02/20/2024. To ensure all relevant diagnoses for this study were included; three separate keyword searches were performed using "glioma", "astrocytoma", and "glioblastoma". The search only included keyword results that were present in the Final Diagnoses. "Glioma" returned 85 cases; "Astrocytoma" returned 67 cases; and "Glioblastoma" returned 215 cases. Following the exclusion of duplicate cases, those missing any of the four requisite MR imaging sequences, and cases that failed processing through our pipeline, our final cohort comprised 203 patients.

Data Acquisition

Radiology: MRI studies on our McKesson Radiology 12.2 Picture archiving and communication system (PACS) (Change Healthcare Radiology Solutions, Nashville, Tennessee, U.S) were exported. The image exportation process involved multiple personnels of varying ranks, including medical graduates, radiology residents, neuroradiology fellows, and neuroradiologists. Our team exported the four basic conventional MR sequences including T1, T1 with IV gadolinium-based contrast agent administration, T2, and Fluid Attenuated Inversion Recovery (FLAIR) into a HIPPA compliant MU secured research server.

For each patient, the images were thoroughly checked for including up to six post-treatment images as available. The post-treatment images were captured on different dates, though not all patients had the maximum number of follow-up images; some had as few as one post-treatment follow-up MRI. For patients with more frequent follow-up MRIs, the immediate post-operative scan, at least one time point of progression and another follow-up study. The MR images were comprehensively reviewed to exclude significantly motion degraded or suboptimal studies.

The majority of the studies were conducted using Siemens MRI machines 97.47%, n=579 with a smaller proportion performed on MRI machines from other vendors: GE (2.02%, n=12) and Philips (0.51%, n=3). Table 1 shows the distribution of studies across different Siemens MR machines. Regarding the magnetic field strength, 1.5T MRIs accounted for 48.14% (n=1,126), 3T MRIs accounted for 45.08% (n=318), and 3T MRIs accounted for 45.08% (n=261). Table 2 summarizes the MRI parameters of each MR sequence.

Our team made efforts to obtain 3D sequences whenever available. Scans were performed using 3D acquisition methods in 40.28% of cases (n=975) and 2D acquisition methods in 59.82% of cases (n=1,419). In cases where 3D images were not available, 2D images were utilized instead. Table 3 summarizes the counts and percentage of studies performed with 2D vs 3D acquisition across different MR sequences.

Clinical: Basic demographic data, clinical data points, and tumor pathology were obtained through review of the electronic medical record (EMR). Clinical data points included the date of diagnosis, date of first surgery or treatment, date and characterization of first and/or subsequent disease progression and/or recurrence, and date of any follow-up resections. Survival information included the date of death and, if that was unknown, the date of last known contact while alive. Disease progression and/or recurrence was characterized as imaging only, clinical only, or both based on information obtained through review of each patient’s clinical notes, brain imaging, and clinical impression as documented by the primary care team. Brief summaries of the reasoning behind each characterization were also included. Patients with no further clinical contact beyond their primary treatment were documented as “lost to follow-up.” Pathological information was obtained through review of the initial pathology note and any subsequent addenda for each tumor sample and included final tumor diagnosis, grade, and any identified genetic mutations. This information was then compiled into a spreadsheet for analysis.

Data Analysis

The image data underwent preprocessing using the Federated Tumor Segmentation (FeTS) tool. The pipeline began with converting DICOM files to the Neuroimaging Informatics Technology Initiative (NIfTI) format, ensuring the removal of any remaining PHI not eliminated by the anonymization/de-identification tool. The converted NIfTI images were then resampled to an isotropic 1mm³ resolution and co-registered to the standard anatomical human brain atlas, SRI24. A deep learning brain extraction method was applied to strip the skull and extracranial tissues, thereby mitigating any potential facial reconstruction or recognition risks.

The preprocessed images were segmented using a deep network based on nnU-Net, resulting in four distinct labels that correspond to different components of each tumor:

• Label 1: Non-enhancing Tumor Core (NETC). This label identifies non-enhancing components within the tumor, such as cystic, necrotic, or hemorrhagic portions.
• Label 2: Surrounding Non-enhancing FLAIR Hyperintensity (SNFH). This label represents both non-enhancing infiltrative tumor components and peritumoral vasogenic edema.
• Label 3: Enhancing Tissue (ET). This label highlights the viable nodular-enhancing components of the tumor.
• Label 4: Resection Cavity (RC). This label covers post-surgical changes, including recent changes like blood products and air foci, as well as chronic changes with materials isointense to CSF signal.

A spreadsheet is also provided that includes tumor volumes and signal intensity of different tumor components across various MR sequences.

Usage Notes

Each scan was manually exported using the built-in McKesson DICOM export tool into separate folders labeled as post-treatment 1, post-treatment 2, etc. In a subsequent step, a subset of the data was selected to contribute for the development of FeTS 2 toolbox. Consequently, the naming convention was updated to replace "post-treatment" with "timepoint" (e.g., post-treatment 1 became timepoint 1) to adhere to the instructions of the FeTS development team. Each sequence was saved in its own folder within these categories to a HIPPA compliant and secured server within the University of Missouri network. Exportation was conducted in DICOM format, maintaining the original image compression settings to preserve quality. To ensure patient privacy and HIPPA compliance, all images were anonymized and all protected health information (PHI) e.g. patient name, MRN, accession number, etc. were deleted from the metadata DICOM headers.

The folders are labeled in the following structure:

• Main folder: PatientID_XXXX
• Subfolders: Timepoint_X, Timepoint_X
• Each time point folder has the NIfTI images associated with the respective timepoints.

This dataset contains image annotations derived from the NCI Clinical Trial "Combination Chemotherapy and Radiation Therapy in Treating Young Patients With Newly Diagnosed Hodgkin Lymphoma (AHOD0831)”. This dataset was generated as part of an NCI project to augment TCIA datasets with annotations that will improve their value for cancer researchers and AI developers.

Annotation Protocol

For each patient, all scans were reviewed to identify and annotate the clinically relevant time points and sequences/series. Scans were initially annotated by an international team of radiologists holding MBBS degrees or higher, which were then reviewed by US-based board-certified radiologists to ensure accuracy. In a typical patient all available time points were annotated: In a typical patient the following annotation rules were followed:

1. PERCIST criteria was followed. Specifically, the lesions estimated to have the most elevated SUVmax were annotated.
2. Lesions were annotated in the axial plane. If no axial plane was available, lesions were annotated in the coronal plane.
3. Lesions were annotated on the attenuation corrected PET series as well as the diagnostic contrast enhanced CT. If no diagnostic contrast enhanced CT is available for that timepoint, then the non contrast CT portion of the PET/CT was annotated.
4. A maximum of 5 lesions were annotated per patient scan (timepoint); no more than 2 per organ. For the purposes of this project, the lymph nodes constitute 1 organ, while other lymphatic structures such as the spleen, salivary glands, and Waldeyer’s ring structures constitute separate organs. The same 5 lesions were annotated at each time point. RECIST 1.1 principles will be followed. Specifically, lymph nodes were annotated if > 1.5 cm in short axis. Other lesions were annotated if > 1 cm.
5. Lesions were labeled separately.
6. Seed points were automatically generated but reviewed by a radiologist.
7. To ensure a high standard of accuracy and data quality, each annotation was reviewed by a secondary reader.

At each time point:

1. A seed point (kernel) was created for each segmented structure. The seed points for each segmentation are provided in a separate DICOM RTSS file.
2. SNOMED-CT “Anatomic Region Sequence” and “Segmented Property Category Code Sequence” and codes were inserted for all segmented structures.
3. “Tracking ID” and “Tracking UID” tags were inserted for each segmented structure to enable longitudinal lesion tracking.
4. Imaging time point codes were inserted to help identify each annotation in the context of the clinical trial assessment protocol.
   1. “Clinical Trial Time Point ID” was used to encode time point type using one of the following strings as applicable: “pre-dose” or “post-chemotherapy”
   2. Content Item in “Acquisition Context Sequence” was added containing "Time Point Type" using Concept Code Sequence (0040,A168) selected from:
      1. (255235001, SCT, “Pre-dose”)
      2. (262502001, SCT, "Post-chemotherapy")

Important supplementary information and sample code

1. A spreadsheet containing key details about the annotations is available in the Data Access section below.
2. A Jupyter notebook demonstrating how to use the NBIA Data Retriever Command-Line Interface application and the REST API (with authentication) to access these data can be found in the Additional Resources section below.

Snapshot img

Teleradiology Market Size 2024-2028

The teleradiology market size is forecast to increase by USD 3.83 billion, at a CAGR of 17.55% between 2023 and 2028.

The market is witnessing significant growth, driven by the increasing prevalence of diseases and the growing geriatric population. This demographic shift, coupled with advancements in technology, is fueling the demand for remote diagnostic solutions. Furthermore, the market is experiencing an uptick in collaborations and new product launches, as industry players seek to expand their offerings and cater to the evolving needs of healthcare providers and patients. However, challenges persist, including concerns related to the lack of early diagnosis due to image quality and interpretation issues. Ensuring accurate and timely diagnoses remains a critical priority for market participants, necessitating ongoing investments in technology and training. As the market continues to evolve, companies must navigate these challenges and capitalize on opportunities to deliver high-quality, efficient, and accessible diagnostic services.

What will be the Size of the Teleradiology Market during the forecast period?

Explore in-depth regional segment analysis with market size data - historical 2018-2022 and forecasts 2024-2028 - in the full report.
Request Free SampleThe market continues to evolve, driven by advancements in technology and the increasing demand for remote diagnostic services. High-resolution image viewing, regulatory compliance, image quality assessment, electronic health records, and various diagnostic imaging modalities are seamlessly integrated into the teleradiology workflow. Pacs integration and telemedicine integration enable radiologists to access and interpret images from multiple locations, enhancing efficiency and improving patient care. Network bandwidth requirements and data privacy regulations pose challenges, necessitating the adoption of image compression techniques, secure data transmission protocols, and cybersecurity measures. AI-powered image analysis and clinical decision support tools contribute to more accurate diagnoses and radiation dose optimization. Remote collaboration tools and cloud-based image storage facilitate seamless communication and access to medical images, enabling radiologists to work together and share expertise. Continuous innovation in image annotation tools, DICOM image transfer, and video conferencing technologies further enhance the teleradiology landscape, ensuring that the market remains dynamic and responsive to the evolving needs of the healthcare industry. The ongoing integration of these technologies and regulatory compliance ensures that teleradiology services remain a valuable and essential component of modern healthcare delivery.

How is this Teleradiology Industry segmented?

The teleradiology industry research report provides comprehensive data (region-wise segment analysis), with forecasts and estimates in 'USD million' for the period 2024-2028, as well as historical data from 2018-2022 for the following segments. ModalityCTX-rayUltrasoundMRINuclear imagingComponentHardwareSoftwareTelecom and networkingEnd UseHospitalRadiology ClinicsAmbulatory Imaging CenterHospitalRadiology ClinicsAmbulatory Imaging CenterGeographyNorth AmericaUSCanadaEuropeGermanyUKAPACChinaRest of World (ROW)

By Modality Insights

The ct segment is estimated to witness significant growth during the forecast period.In the dynamic the market, CT scanners play a pivotal role in delivering high-quality diagnostic imaging remotely. These scanners, available in various modalities, each offer unique advantages for teleradiology services. Conventional CT scanners, utilizing X-ray beams, create precise cross-sectional images of the body, making them essential for evaluating conditions like trauma, cancer, and cardiovascular diseases. Widely used for routine diagnostic imaging, they remain the primary modality in teleradiology. The multi-detector CT (MDCT) scanner is another significant modality, employing multiple detector rows to acquire images more swiftly and with higher resolution. This expedited image acquisition process enhances diagnostic accuracy and efficiency in teleconsultation platforms. HIPAA compliance ensures secure data transmission and patient privacy during remote image interpretation. Clinical decision support and image annotation tools facilitate radiologist workflow optimization. Network bandwidth requirements and data privacy regulations necessitate robust image compression techniques and secure image sharing protocols. Cybersecurity protocols and AI-powered image analysis further enhance the security and diagnostic capabilities of teleradiology. High-resolution image viewing and video conferencing enable real-time collaboration between radiologists and healthcare providers. Integration of telemedicine, PACS workflow, and electronic health records streamlines the dia

PurposeTo investigate the inter-reader agreement of using the American College of Radiology (ACR) Thyroid Imaging Reporting and Data System (TI-RADS) for risk stratification of thyroid nodules.MethodsA literature search of Web of Science, PubMed, Cochrane Library, EMBASE, and Google Scholar was performed to identify eligible articles published from inception until October 31, 2021. We included studies reporting inter-reader agreement of different radiologists who applied ACR TI-RADS for the classification of thyroid nodules. Quality assessment of the included studies was performed with the Quality Assessment of Diagnostic Accuracy Studies-2 tool and Guidelines for Reporting Reliability and Agreement Studies. The summary estimates of the inter-reader agreement were pooled with the random-effects model, and multiple subgroup analyses and meta-regression were performed to investigate various clinical settings.ResultsA total of 13 studies comprising 5,238 nodules were included in the current meta-analysis and systematic review. The pooled inter-reader agreement for overall ACR TI-RADS classification was moderate (κ = 0.51, 95% CI 0.42–0.59). Substantial heterogeneity was presented throughout the studies, and meta-regression analyses suggested that the malignant rate was the significant factor. Regarding the ultrasound (US) features, the best inter-reader agreement was composition (κ = 0.58, 95% CI 0.53–0.63), followed by shape (κ = 0.57, 95% CI 0.41–0.72), echogenicity (κ = 0.50, 95% CI 0.40–0.60), echogenic foci (κ = 0.44, 95% CI 0.36–0.53), and margin (κ = 0.34, 95% CI 0.24–0.44).ConclusionsThe ACR TI-RADS demonstrated moderate inter-reader agreement between radiologists for the overall classification. However, the US feature of margin only showed fair inter-reader reliability among different observers.

Snapshot img

Online Radiology Education Platforms Market Size 2025-2029

The global online radiology education platforms market is anticipated to grow by USD 1.15 billion, exhibiting a CAGR of approximately 7.3% during the forecast period. Key factors driving this expansion include technological advancements, flexible learning options, and the increasing emphasis on continuous professional development. The global online radiology education platforms market is experiencing growth driven by the increasing demand for specialized training and the need for accessible learning options. These platforms offer unparalleled convenience for healthcare professionals with busy schedules, enabling radiologists to access expert-led case reviews and lectures in short, accessible formats from any location. Modern platforms are combining virtual reality simulations, 3D anatomical models, real-time case studies, interactive assessment tools, and peer collaboration features, catering to different learning styles and preferences, thereby boosting market expansion.

To access the full market forecast and comprehensive analysis, Buy Now

How is this market segmented?

The market research report provides comprehensive data (region-wise segment analysis), with forecasts and estimates in USD bn for the period 2025-2029, as well as historical data from 2019-2024 for the following segments:

Component

  Software Solutions
  Services


Delivery Mode

  Cloud/Web-Based
  On-Premises


Mode Of Learning

  Self-Paced Learning
  Instructor-Led Training
  Blended Learning (Hybrid Model)


End Use

  Medical Students & Residents
  Radiologists & Physicians
  Hospitals & Imaging Centers


Application

  Diagnostic Radiology
  Oncology Imaging
  Interventional Radiology


Software Subscription Type

  Monthly/Annual Subscription
  Freemium Model with Paid Certifications


APAC

  China
  India
  Japan
  Australia
  Rest of APAC

Europe


  Germany
  Spain
  Italy
  UK
  Rest of Europe

North America


  US
  Canada

South America & MEA


  Brazil
  UAE
  South Africa
  Others





End Use

  Radiologists & Physicians: Fueled by the increasing need for continuous professional development and staying current with evolving diagnostic techniques.
  Hospitals & Imaging Centers: Due to increasing investments to integrate advanced training programs
  Medical Students & Residents: Driven by the need for foundational knowledge and practical skills in radiology education, with a focus on early exposure and flexible study options.


Component

  Services: Fueled by the growing need for personalized, expert-led training, including live webinars, mentorship, and tailored consulting services.
  Software Solutions: Driven by increasing demand for digital learning and advanced educational technologies, including interactive learning tools and AI-driven assessments.


Application

  Oncology Imaging: Fueled by the rising incidence of cancer and advancements in imaging technology for precision medicine and targeted therapies.
  Diagnostic Radiology: Driven by increasing demand for specialized training in diagnostic imaging techniques amid growing complexity of imaging modalities.
  Interventional Radiology: Due to increasing need for real time practical guidance to the physicians


Delivery Mode

  On-Premises: Driven by need for enhanced security, control, and customization, especially among healthcare institutions safeguarding sensitive data.
  Cloud/Web-Based: Growth driven by increasing demand for remote learning, ease of integration with existing healthcare systems, and ability to deliver interactive content.


Mode Of Learning

  Self-Paced Learning: Driven by radiologists' need for flexibility, enabling them to access content at their own pace without disrupting work schedules.
  Instructor-Led Training: Driven by the increasing demand for expert guidance, real-time feedback, and interactive learning experiences.
  Blended Learning (Hybrid Model): Due to its capability to offer best of both worlds, the Blended learning method is expected to gain traction


Software Subscription Type

  Monthly/Annual Subscription: Driven by its cost-effectiveness, convenience, and continuous access to educational content.
  Freemium Model with Paid Certifications: Driven by its ability to attract a broad user base while offering advanced learning and certification options, incentivizing professionals to upgrade skills.

Regional Analysis

APAC: The Asia Pacific region is gaining traction due to increasing demand for skilled radiologists and the rise of telemedicine. There is integration of advanced technologies, including AI and machine learning, to personalize learning experiences and improve diagnostic accuracy. Strategic partnerships help to expand access to advanced imaging technologies and educational resources.
Europe: The Europe online radiology education platf

This dataset contains image annotations derived from the NCI Clinical Trial "Chemotherapy and Radiation Therapy in Treating Young Patients With Newly Diagnosed, Previously Untreated, High-Risk Medulloblastoma/PNET (ACNS0332)". This curated dataset provides a comprehensive picture of imaging in pediatric patients with newly diagnosed primitive neuroectodermal tumors throughout their treatment and until any potential relapse. This is the largest known dataset of patients with supratentorial primitive neuroectodermal tumors and pineoblastomas. This dataset was generated as part of an NCI project to augment TCIA datasets with annotations that will improve their value for cancer researchers and AI developers.

Annotation Protocol

For each patient, all scans were reviewed to identify and annotate the clinically relevant time points and sequences/series. Scans were initially annotated by an international team of radiologists holding MBBS degrees or higher, which were then reviewed by US-based board-certified radiologists to ensure accuracy. In a typical patient the following time points were annotated:

1. Pre-surgical study
2. Post-surgical study [if applicable]
3. Follow-up study at the completion of radiotherapy.
4. Follow-up study at the end of chemotherapy.
5. Follow-up study relapse [if applicable]

At each time point, 3D segmentations (DICOM SEG), seed points (DICOM RTSTRUCT) and negative finding assessments (DICOM RTSTRUCT) were created:

1. Enhancing tumor on an axial 3D T1 post contrast sequence
   1. If not available, a 3D post contrast sequence in another plane was used.
   2. If no 3D post contrast sequence was available, the tumor was annotated in all 3 planes utilizing 2D post contrast sequences.
   3. On post-contrast sequences, the entire tumor, including the cystic and non enhancing components was annotated.
   4. Any resection cavity or post-op changes/products was excluded.
2. Edema on an axial T2 FLAIR sequence
   1. If not available, an axial T2 or other T2 weighted sequence was used.
   2. The segmentation mask contains both the edematous tissue and the tumor.
3. The portion of the tumor demonstrating restricted diffusion on an ADC sequence
4. Up to 5 metastatic lesions within the brain and and up to 5 metastatic lesions in the spine as demonstrated on whatever T1 post contrast sequence they are visualized on
   1. When present, the 5 largest lesions were annotated.
5. A manually placed seed point (kernel) were created for each segmented structure
   1. The seed points for each segmentation are provided in a separate DICOM RTSTRUCT file.
   2. Spinal metastases, which are too small to apply a volumetric mask to, only have a seed point annotation.
6. If no seed points or segmentations were generated a "Negative Assessment Report" RTSTRUCT file was created to record this fact.
7. To ensure a high standard of accuracy and data quality, each annotation was reviewed by a secondary reader.

Important supplementary information and sample code

1. A spreadsheet containing a variety of useful metadata about the annotations, including calculated tumor volumes, is available in the Data Access section below.
2. Important information about how to interpret the DICOM annotation data can be found on the Detailed Description section below. It includes information about specific tags which document where the tumor was found, whether it was enhancing/non-enhancing, which study time point the annotation relates to, details for lesion tracking across time points, etc.
3. A Jupyter notebook demonstrating how to use the NBIA Data Retriever Command-Line Interface application and our REST API (with authentication) to access these data can be found in the Additional Resources section below.
4. Instructions for visualizing these data in 3D Slicer can be found in the Additional Resources section below.

This dataset contains image annotations derived from the NCI Clinical Trial "A Randomized Phase III Study Comparing Carboplatin/Paclitaxel or Carboplatin/Paclitaxel/Bevacizumab With or Without Concurrent Cetuximab in Patients With Advanced Non-small Cell Lung Cancer”. This dataset was generated as part of an NCI project to augment TCIA datasets with annotations that will improve their value for cancer researchers and AI developers.

Annotation Protocol

For each patient, all scans were reviewed to identify and annotate the clinically relevant time points and sequences/series. Scans were initially annotated by an international team of radiologists holding MBBS degrees or higher, which were then reviewed by US-based board-certified radiologists to ensure accuracy. In a typical patient, all available time points were annotated. Every exam from the first available time point was annotated. One additional time point was annotated for each patient. The clinical data in the NCTN Archive was utilized to help determine the first evidence of disease progression. The first time point demonstrating disease progression was annotated. If that document was not accurate and did not demonstrate disease progression, then later time points were reviewed to assess for disease progression and the first time point demonstrating disease progression was annotated. If there was no evidence of disease progression on any time point, then the last available time point was annotated. Again, every exam from each chosen time point was annotated. For example, if there was a CT and a PET/CT, the PET was annotated along with one CT. If there was an MRI, that was annotated as well. The following annotation rules were followed:

1. PERCIST criteria was followed for PET imaging. Specifically, the lesions estimated to have the most elevated SUVmax were annotated.
2. RECIST 1.1 was otherwise generally followed for any MR and CT imaging. A maximum of 5 lesions were annotated per patient scan (timepoint); no more than 2 per organ. The same 5 lesions were annotated at each time point. Lymph nodes were however annotated if > 1 cm in short axis. Other lesions were annotated if > 1 cm. If the primary lesion is < 1 cm, it was still annotated. If there was evidence of disease progression with new lesions then additional annotations were allowed to demonstrate that progression. A representative sample of the new lesions was annotated at the radiologist's discretion.
3. Lesions were annotated in the axial plane. If no axial plane was available, lesions were annotated in the coronal plane.
4. MRIs were annotated using the T1-weighted post contrast sequence, fat saturated if available. If not available, T2-weighted sequences were utilized.
5. CTs were annotated using the axial post contrast series. If not available, the non contrast series was annotated.
6. PET/CTs were annotated on the CT and attenuation corrected PET images. However, if the post contrast CT was performed the same day as the PET/CT, the non contrast CT portion of the PET/CT was annotated.
7. Lesions were labeled separately.
8. The volume of each annotated lesion was calculated and reported in cubic centimeters [cc] in a dataset metadata report.
9. Seed points were automatically generated but reviewed by a radiologist.
10. A “negative” annotation was created for any exam without findings.

At each time point:

1. A seed point (kernel) was created for each segmented structure. The seed points for each segmentation are provided in a separate DICOM RTSS file.
2. SNOMED-CT “Anatomic Region Sequence” and “Segmented Property Category Code Sequence” and codes were inserted for all segmented structures.
3. “Tracking ID” and “Tracking UID” tags were inserted for each segmented structure to enable longitudinal lesion tracking.
4. Imaging time point codes were inserted to help identify each annotation in the context of the clinical trial assessment protocol.
   1. “Clinical Trial Time Point ID” was used to encode time point type using one of the following strings as applicable: “pre-dose” or “post-chemotherapy”
   2. Content Item in “Acquisition Context Sequence” was added containing "Time Point Type" using Concept Code Sequence (0040,A168) selected from:
      1. (255235001, SCT, “Pre-dose”)
      2. (262502001, SCT, "Post-chemotherapy")

Important supplementary information and sample code

1. A spreadsheet containing key details about the annotations is available in the Data Access section below.
2. A Jupyter notebook demonstrating how to use the NBIA Data Retriever Command-Line Interface application and the REST API to access these data can be found in the Additional Resources section below.

This dataset contains image annotations derived from the NCI Clinical Trial "Combination Chemotherapy and Surgery in Treating Young Patients With Wilms Tumor (AREN0534)”. This dataset was generated as part of an NCI project to augment TCIA datasets with annotations that will improve their value for cancer researchers and AI developers.

Annotation Protocol

For each patient, all scans were reviewed to identify and annotate the clinically relevant time points and sequences/series. Scans were initially annotated by an international team of radiologists holding MBBS degrees or higher, which were then reviewed by US-based board-certified radiologists to ensure accuracy. In a typical patient the following time points were annotated:

1. Pre-surgical CT chest and CT/MRI abdomen
2. CT chest and/or CT/MRI abdomen at 6 weeks
3. Possible CT/MRI abdomen at 12 weeks.
4. Any negative imaging included past 12 weeks was annotated as negative. If any included imaging past 12 weeks is positive for tumor, the last positive exam was annotated.

In a typical patient the following annotation rules were followed:

1. The primary renal tumor(s) were annotated on post-contrast axial series. Normal renal parenchyma were excluded.
2. A maximum of 5 lesions were annotated per patient scan (timepoint); no more than 2 per organ. The same 5 lesions were annotated at each time point. RECIST 1.1 principles were followed for lesion annotation, however, if <5 lesions measuring >1 cm were present, then smaller lesions were annotated, again up to 2 lesions per organ or 5 lesions per patient scan. Bone lesions were included if other lesions were not present.
3. Lesions were labeled separately.
4. Seed points were automatically generated but reviewed by a radiologist.
5. To ensure a high standard of accuracy and data quality, each annotation was reviewed by a secondary reader.

At each time point:

1. A seed point (kernel) was created for each segmented structure. The seed points for each segmentation are provided in a separate DICOM RTSS file.
2. SNOMED-CT “Anatomic Region Sequence” and “Segmented Property Category Code Sequence” and codes were inserted for all segmented structures.
3. Imaging time point codes were inserted to help identify each annotation in the context of the clinical trial assessment protocol.
   1. “Clinical Trial Time Point ID” was used to encode time point type using one of the following strings as applicable: “pre-dose”, “post-chemotherapy”, or “post-operative”.
   2. Content Item in “Acquisition Context Sequence” will be added containing "Time Point Type" using Concept Code Sequence (0040,A168) selected from:
      1. (255235001, SCT, “Pre-dose”)
      2. (262502001, SCT, "Post-chemotherapy")
      3. (262061000, SCT, "Post-operative")

Important supplementary information and sample code

1. A spreadsheet containing key details about the annotations is available in the Data Access section below.
2. A Jupyter notebook demonstrating how to use the NBIA Data Retriever Command-Line Interface application and the REST API (with authentication) to access these data can be found in the Additional Resources section below.

This dataset contains image annotations derived from the NCI Clinical Trial "ACRIN-HNSCC-FDG-PET-CT (ACRIN 6685)”. This dataset was generated as part of an NCI project to augment TCIA datasets with annotations that will improve their value for cancer researchers and AI developers.

Annotation Protocol

For each patient, all scans were reviewed to identify and annotate the clinically relevant time points and sequences/series. Scans were initially annotated by an international team of radiologists holding MBBS degrees or higher, which were then reviewed by US-based board-certified radiologists to ensure accuracy. In a typical patient all available time points were annotated. The following annotation rules were followed:

1. PERCIST criteria was followed for PET imaging. Specifically, the lesions estimated to have the most elevated SUVmax were annotated.
2. RECIST 1.1 was otherwise generally followed for MR and CT imaging. A maximum of 5 lesions were annotated per patient scan (timepoint); no more than 2 per organ. The same 5 lesions were annotated at each time point. Lymph nodes were annotated if >1 cm in short axis. Other lesions were annotated if >1 cm. If the primary lesion is < 1 cm, it was still annotated.
3. Three-dimensional segmentations of lesions were created in the axial plane. If no axial plane was available, lesions were annotated in the coronal plane.
4. MRIs were annotated using the T1-weighted axial post contrast sequence, fat saturated if available.
5. CTs were annotated using the axial post contrast series. If not available, the non contrast series were annotated.
6. PET/CTs were annotated on the CT and attenuation corrected PET images.
7. If the post contrast CT was performed the same day as the PET/CT, the non contrast CT portion of the PET/CT was not annotated.
8. Lesions were labeled separately.
9. The volume of each annotated lesion was calculated and reported in cubic centimeters [cc] in the Annotation Metadata CSV.
10. Seed points were automatically generated, but reviewed by a radiologist.
11. A “negative” annotation was created for any exam without findings.

At each time point:

1. A seed point (kernel) was created for each segmented structure. The seed points for each segmentation are provided in a separate DICOM RTSTRUCT file.
2. SNOMED-CT “Anatomic Region Sequence” and “Segmented Property Category Code Sequence” and codes were inserted for all segmented structures.
3. “Tracking ID” and “Tracking UID” tags were inserted for each segmented structure to enable longitudinal lesion tracking.
4. Imaging time point codes were inserted to help identify each annotation in the context of the clinical trial assessment protocol.
   1. “Clinical Trial Time Point ID” was used to encode time point type using one of the following strings as applicable: “pre-dose” or “post-chemotherapy”.
   2. Content Item in “Acquisition Context Sequence” was added containing "Time Point Type" using Concept Code Sequence (0040,A168) selected from:
      1. (255235001, SCT, “Pre-dose”) (in this trial, both the CT/MRI and PET/CT, while being different timepoints, are pre-treatment)

Important supplementary information and sample code

1. A spreadsheet containing key details about the annotations is available in the Data Access section below.
2. A Jupyter notebook demonstrating how to use the NBIA Data Retriever Command-Line Interface application and the REST API to access these data can be found in the Additional Resources section below.

CheXpert

CheXpert is a large dataset of chest X-rays and competition for automated chest x-ray interpretation, which features uncertainty labels and radiologist-labeled reference standard evaluation sets. https://stanfordmlgroup.github.io/competitions/chexpert/

  Warning on AP/PA label

I could not find in the paper a mapping from the 0/1 label to AP/PA, so I assumed 0=AP and 1=PA. Looking at a few images this seems to be correct, but I'm not a radiologist.… See the full description on the dataset page: https://huggingface.co/datasets/danjacobellis/chexpert.

Snapshot img

AI In Medical Imaging Market Size 2025-2029

The AI in medical imaging market size is forecast to increase by USD 5.17 billion at a CAGR of 28.1% between 2024 and 2029.

The market is experiencing significant growth, driven by the escalating demand for rapid and precise diagnostics. This need for efficiency and accuracy is particularly prominent in the healthcare sector, where timely and accurate diagnoses can significantly impact patient outcomes. A key trend shaping this market is the ascendance of generative AI in reshaping medical diagnostics. Generative AI, a subset of artificial intelligence, has the ability to create new and unique content, making it an invaluable tool in medical imaging. It can analyze vast amounts of data to identify patterns and make accurate diagnoses, often surpassing the capabilities of human experts.
However, the market is not without challenges. Navigating the complexities of data privacy and security is a significant obstacle. With the increasing use of AI in medical imaging comes the need to protect sensitive patient information. Ensuring data security and privacy is crucial to maintain trust and confidence in the technology. Companies must invest in robust security measures and adhere to strict regulations to mitigate risks and safeguard patient data. Training programs and continuing medical education ensure professionals stay updated with the latest techniques and ethical considerations.

What will be the Size of the AI In Medical Imaging Market during the forecast period?

Explore in-depth regional segment analysis with market size data - historical 2019-2023 and forecasts 2025-2029 - in the full report.
Request Free Sample

The market for AI in medical imaging continues to evolve, with applications spanning various sectors including high-throughput screening, biomarker discovery tools, and personalized medicine strategies. Regulatory approval processes are a critical component of this dynamic market, ensuring the safety and efficacy of AI-powered medical imaging solutions. For instance, a recent study demonstrated a 25% increase in accuracy for tumor volume quantification using AI-assisted image analysis, compared to traditional methods. Moreover, AI is revolutionizing microscopy image processing, medical image classification, and pathology image analysis, enabling real-time results and improving clinical decision-making. Performance benchmarking studies and data visualization dashboards facilitate workflow optimization strategies, while data security measures ensure ethical considerations are met. Advanced image enhancement technologies, such as deep learning and artificial intelligence (AI), are revolutionizing diagnostic imaging.

In the realm of treatment planning software, AI-driven algorithms are increasingly being used for lesion detection, radiology report generation, and clinical decision support. Cost-effectiveness evaluation, interoperability standards, and physician training programs further contribute to the ongoing unfolding of market activities. The market is expected to grow by over 20% annually, driven by the continuous development of advanced AI technologies and their integration into various medical applications. Data security and clinical workflow optimization are critical considerations for healthcare organizations implementing medical imaging solutions, with PACS systems and data security measures playing a vital role in ensuring patient privacy and data integrity.

How is this AI In Medical Imaging Industry segmented?

The AI in medical imaging industry research report provides comprehensive data (region-wise segment analysis), with forecasts and estimates in 'USD million' for the period 2025-2029, as well as historical data from 2019-2023 for the following segments.

End-user

  Hospitals
  Diagnostic imaging centers
  Others


Application

  Radiology
  Cardiology
  Neurology
  Oncology
  Others


Type

  CT scan
  X-ray
  MRI
  Ultrasound
  Others


Geography

  North America

    US
    Canada


  Europe

    France
    Germany
    Italy
    Spain
    UK


  APAC

    China
    India
    Japan


  Rest of World (ROW)

By End-user Insights

The Hospitals segment is estimated to witness significant growth during the forecast period. The artificial intelligence (AI) in medical imaging market is witnessing significant growth, driven by the adoption of advanced technologies such as transfer learning models, cloud-based image processing, and patient-specific modeling. Hospitals, with their substantial financial resources, sophisticated IT infrastructure, and extensive patient datasets, lead the market with a share of over 64%. AI's implementation in hospitals is primarily motivated by the need to enhance diagnostic accuracy, increase operational efficiency, and address the global shortage of radiologists. By addressing these challenges and capitali

Snapshot img

AI In MRI Market Size 2025-2029

The AI in MRI market size is forecast to increase by USD 1.03 billion, at a CAGR of 27.8% between 2024 and 2029.

The market is experiencing significant growth, driven by increasing pressure on radiology departments and workforce shortages. Hospitals and clinics are turning to AI solutions to streamline processes, improve efficiency, and enhance diagnostic accuracy. The integration of AI platforms and digital marketplaces is a key trend, enabling seamless access to advanced imaging technologies and expert radiology services. However, challenges persist, including data quality, generalizability, and privacy concerns. Data analytics plays a crucial role in optimizing clinical workflow and improving patient experience.
Companies seeking to capitalize on market opportunities must focus on addressing these challenges and delivering solutions that prioritize data security, transparency, and regulatory compliance. By doing so, they will be well-positioned to meet the evolving needs of healthcare providers and improve patient outcomes. Ensuring the accuracy and reliability of AI algorithms is crucial, as is addressing the ethical implications of handling sensitive patient data.

What will be the Size of the AI In MRI Market during the forecast period?

Get Key Insights on Market Forecast (PDF) Request Free Sample

The market continues to evolve, with advancements in technologies such as 3D MRI reconstruction, pattern recognition MRI, and radiomics MRI features revolutionizing diagnostic accuracy and efficiency. MRI image denoising and noise reduction techniques enhance image quality, while fast MRI acquisition and artifact correction streamline workflows. Quantitative MRI and brain MRI segmentation enable more precise diagnoses, and AI-powered MRI diagnostics offer faster, more reliable results. Real-time MRI processing and functional MRI analysis provide new insights into patient conditions. Machine learning MRI and AI-driven MRI workflows optimize image reconstruction and feature extraction.
Predictive MRI modeling and AI-driven MRI protocols further enhance diagnostic capabilities. For instance, a leading research institution reported a 25% increase in lesion detection accuracy using AI-assisted MRI interpretation. The MRI industry anticipates a 15% compound annual growth rate in the coming years, driven by these technological advancements. The integration of advanced technologies like artificial intelligence (AI) and machine learning is expected to enhance the capabilities of MRI systems further.

How is this AI In MRI Industry segmented?

The AI in MRI industry research report provides comprehensive data (region-wise segment analysis), with forecasts and estimates in 'USD million' for the period 2025-2029, as well as historical data from 2019-2023 for the following segments.

Component

  Software
  Services
  Hardware


End-user

  Hospitals
  Diagnostic imaging centers
  Others


Application

  Neurology
  Musculoskeletal
  Cardiovascular
  Prostate
  Others


Geography

  North America

    US
    Canada
    Mexico


  Europe

    France
    Germany
    Italy
    UK


  APAC

    China
    Japan


  South America

    Brazil


  Rest of World (ROW)

By Component Insights

The Software segment is estimated to witness significant growth during the forecast period. The market is experiencing significant growth, with the software segment leading the innovation. This segment includes solutions from both established medical imaging corporations and specialized AI startups. One sub-segment of note is AI algorithms for image reconstruction, which utilize deep learning to generate high-resolution images from under-sampled data, reducing scan times without sacrificing diagnostic quality. Another crucial software category is computer-aided detection and diagnosis. These platforms assist radiologists by automatically identifying, segmenting, and characterizing potential abnormalities, such as tumors or lesions, enhancing diagnostic accuracy and efficiency.

Tasks like image registration MRI, MRI artifact correction, and MRI feature extraction have been significantly automated using AI tools, reducing manual intervention and errors. In high-resolution imaging, high-resolution MRI combined with anatomical MRI segmentation ensures accurate tissue differentiation and pathological identification. Maintaining the integrity of imaging processes, AI-based MRI quality control systems monitor and correct inconsistencies. These systems are also crucial for MRI lesion detection, which is vital in the early diagnosis of conditions like tumors or multiple sclerosis. Moreover, MRI scan optimization powered by AI reduces scan times and enhances patient comfort.

For instance, a leading AI-powered MRI system reportedly achieves a detection rate of 95% for brain lesions, significantly improving cli

This dataset contains image annotations derived from "The Clinical Proteomic Tumor Analysis Consortium Head and Neck Squamous Cell Carcinoma Collection (CPTAC-HNSCC)”. This dataset was generated as part of a National Cancer Institute project to augment images from The Cancer Imaging Archive with tumor annotations that will improve their value for cancer researchers and artificial intelligence experts.

Annotation Protocol

For each patient, all scans were reviewed to identify and annotate the clinically relevant time points and sequences/series. Scans were initially annotated by an international team of radiologists holding MBBS degrees or higher, which were then reviewed by US-based board-certified radiologists to ensure accuracy. In a typical patient all available time points were annotated. The following annotation rules were followed:

1. PERCIST criteria was followed for PET imaging. Specifically, the lesions estimated to have the most elevated SUVmax were annotated.
2. RECIST 1.1 was otherwise generally followed for MR and CT imaging. A maximum of 5 lesions were annotated per patient scan (timepoint); no more than 2 per organ. The same 5 lesions were annotated at each time point. Lymph nodes were annotated if >1 cm in short axis. Other lesions were annotated if >1 cm. The primary tumor was still annotated if < 1 cm.
3. Three-dimensional segmentations of lesions were created in the axial plane. If no axial plane was available, lesions were annotated in the coronal plane.
4. MRIs were annotated using the T1-weighted axial post contrast sequence.
5. CTs were annotated using all axial post contrast series.
6. PET/CTs were annotated on the CT and attenuation corrected PET images, unless there was a diagnostic CT from the same time point, in which case the CT portion of the PET/CT was not annotated.
7. Lesions were labeled separately.
8. Seed points were automatically generated, but reviewed by a radiologist.
9. A “negative” annotation was created for any exam without findings.

At each time point:

1. Volume calculations were performed for each segmented structure. These calculations are provided in the Annotation Metadata CSV.
2. A seed point (kernel) was created for each segmented structure. The seed points for each segmentation are provided in a separate DICOM RTSTRUCT file.
3. SNOMED-CT “Anatomic Region Sequence” and “Segmented Property Category Code Sequence” and codes were inserted for all segmented structures.
4. Imaging time point codes were inserted to help identify each annotation in the context of the clinical trial assessment protocol.
   1. “Clinical Trial Time Point ID” was used to encode time point type using one of the following strings as applicable: “pre-dose” or “post-chemotherapy”.
   2. Content Item in “Acquisition Context Sequence” was added containing "Time Point Type" using Concept Code Sequence (0040,A168) selected from:
      1. (255235001, SCT, “Pre-dose”)
      2. (719864002, SCT, "Post-cancer treatment monitoring")

Important supplementary information and sample code

1. A spreadsheet containing key details about the annotations is available in the Data Access section below.
2. A Jupyter notebook demonstrating how to use the NBIA Data Retriever Command-Line Interface application and the REST API to access these data can be found in the Additional Resources section below.

This dataset contains image annotations derived from the NCI Clinical Trial "Rituximab and Combination Chemotherapy in Treating Patients With Diffuse Large B-Cell Non-Hodgkin's Lymphoma (CALGB50303)”. This dataset was generated as part of an NCI project to augment TCIA datasets with annotations that will improve their value for cancer researchers and AI developers.

Annotation Protocol

For each patient, all scans were reviewed to identify and annotate the clinically relevant time points and sequences/series. Scans were initially annotated by an international team of radiologists holding MBBS degrees or higher, which were then reviewed by US-based board-certified radiologists to ensure accuracy. In a typical patient all available time points were annotated. In a typical patient the following annotation rules were followed: a. PERCIST criteria was followed. Specifically, the lesions estimated to have the most elevated SUVmax were annotated. b. Lesions were annotated in the axial plane. If no axial plane were available, lesions were annotated in the coronal plane. c. Lesions were annotated on the attenuation corrected PET series as well as the diagnostic contrast-enhanced CT. If no diagnostic contrast-enhanced CT was available for that timepoint, then the non-contrast CT portion of the PET/CT was annotated. d. A maximum of 5 lesions were annotated per patient scan (timepoint); no more than 2 per organ. For the purposes of this project, the lymph nodes constitute 1 organ, while other lymphatic structures such as the spleen, salivary glands, and Waldeyer’s ring structures constitute separate organs. The same 5 lesions were annotated at each time point. RECIST 1.1 principles were followed. Specifically, lymph nodes were annotated if > 1.5 cm in short axis. Other lesions were annotated if > 1 cm. e. Lesions were labeled separately. f. Seed points were automatically generated and reviewed by a radiologist. At each time point:

1. A seed point (kernel) was created for each segmented structure. The seed points for each segmentation are provided in a separate DICOM RTSS file.
2. SNOMED-CT “Anatomic Region Sequence” and “Segmented Property Category Code Sequence” and codes were inserted for all segmented structures.
3. “Tracking ID” and “Tracking UID” tags were inserted for each segmented structure to enable longitudinal lesion tracking.
4. Imaging time point codes were inserted to help identify each annotation in the context of the clinical trial assessment protocol.
   1. “Clinical Trial Time Point ID” was used to encode time point type using one of the following strings as applicable: “pre-dose” or “post-chemotherapy”
   2. Content Item in “Acquisition Context Sequence” was added containing "Time Point Type" using Concept Code Sequence (0040,A168) selected from:
      1. (255235001, SCT, “Pre-dose”)
      2. (262502001, SCT, "Post-chemotherapy")

Important supplementary information and sample code

1. A spreadsheet containing key details about the annotations is available in the Data Access section below.
2. A Jupyter notebook demonstrating how to use the NBIA Data Retriever Command-Line Interface application and the REST API (with authentication) to access these data can be found in the Additional Resources section below.

This dataset contains image annotations derived from "The Clinical Proteomic Tumor Analysis Consortium Pancreatic Ductal Adenocarcinoma Collection (CPTAC-PDA)”. This dataset was generated as part of a National Cancer Institute project to augment images from The Cancer Imaging Archive with tumor annotations that will improve their value for cancer researchers and artificial intelligence experts.

Annotation Protocol

For each patient, all scans were reviewed to identify and annotate the clinically relevant time points and sequences/series. Scans were initially annotated by an international team of radiologists holding MBBS degrees or higher, which were then reviewed by US-based board-certified radiologists to ensure accuracy. In a typical patient all available time points were annotated. The following annotation rules were followed:

1. PERCIST criteria was followed for PET imaging. Specifically, the lesions estimated to have the most elevated SUVmax were annotated.
2. RECIST 1.1 was otherwise generally followed for MR and CT imaging. A maximum of 5 lesions were annotated per patient scan (timepoint); no more than 2 per organ. The same 5 lesions were annotated at each time point. Lymph nodes were annotated if >1 cm in short axis. Other lesions were annotated if >1 cm. If the primary lesion measures < 1 cm, it was still annotated.
3. Three-dimensional segmentations of lesions were created in the axial plane. If no axial plane was available, lesions were annotated in the coronal plane.
4. MRIs were annotated using axial T1-weighted post contrast sequences that best demonstrated the tumor.
5. CTs were annotated using all axial post contrast series. If not available, the axial non-contrast series were annotated.
6. PET/CTs were annotated on the CT and attenuation corrected PET images, unless there was a diagnostic CT from the same time point, in which case the CT portion of the PET/CT was not annotated.
7. Lesions were labeled separately.
8. Seed points were automatically generated, but reviewed by a radiologist.
9. A “negative” annotation was created for any exam without findings.

At each time point:

1. Volume calculations were performed for each segmented structure. These calculations are provided in the Annotation Metadata CSV.
2. A seed point (kernel) was created for each segmented structure. The seed points for each segmentation are provided in a separate DICOM RTSTRUCT file.
3. SNOMED-CT “Anatomic Region Sequence” and “Segmented Property Category Code Sequence” and codes were inserted for all segmented structures.
4. Imaging time point codes were inserted to help identify each annotation in the context of the clinical trial assessment protocol.
   1. “Clinical Trial Time Point ID” was used to encode time point type using one of the following strings as applicable: “pre-dose” or “post-chemotherapy”.
   2. Content Item in “Acquisition Context Sequence” was added containing "Time Point Type" using Concept Code Sequence (0040,A168) selected from:
      1. (255235001, SCT, “Pre-dose”)
      2. (719864002, SCT, "Post-cancer treatment monitoring")

Important supplementary information and sample code

1. A spreadsheet containing key details about the annotations is available in the Data Access section below.
2. A Jupyter notebook demonstrating how to use the NBIA Data Retriever Command-Line Interface application and the REST API to access these data can be found in the Additional Resources section below.