Medical Imaging & AI in Oncology
A CT Dataset with RECIST Measurements and Comprehensive Segmentation Masks for Tumors and Lymph Nodes
This research introduces a novel CT imaging dataset crucial for advancing AI in oncology. It includes 1,246 manually segmented lesions across 58 CT scans from 22 cancer patients, complete with RECIST 1.1 compliant diameter measurements. This dataset fills a significant gap in public resources by offering comprehensive lesion annotations (primary tumors, metastases, lymph nodes) and standardized measurements, essential for developing and validating automated RECIST workflows, radiomics studies, and medical imaging foundation models. Its inclusion of Latin American patient data enhances global representation.
Executive Impact at a Glance
Leverage this innovative dataset to accelerate your enterprise's AI capabilities in medical imaging, driving efficiency and precision in oncology.
1246+
Segmented Lesions
58
CT Scans
22
Cancer Patients
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Enterprise Process Flow
CT Image Acquisition
→
Data Anonymization
→
DICOM-to-NIfTI Conversion
→
CT Windowing
→
Data Labeling (Human-in-the-Loop)
→
RECIST Measurements
→
Technical Validation
ρ=0.92
Volume-Diameter Correlation
This significant correlation (Spearman's ρ=0.92) across all lesions highlights the reliability of using diameter length as a proxy for tumor volume in assessing tumor burden, validating a foundational principle for RECIST protocols.
Dataset Strengths vs. Existing Public Datasets
Our dataset addresses critical gaps found in current public datasets, offering unique advantages for AI model development.
| Feature | Existing Public Datasets | This Dataset |
|---|---|---|
| Comprehensive Lesion Annotations | Limited to specific lesion types (e.g., only primary tumors) |
|
| RECIST 1.1 Compliant Measurements | Rarely included or non-standardized |
|
| Instance Segmentation Masks | Often bounding boxes or coarse labels |
|
Accelerating Tumor Response Evaluation with AI
Manual RECIST 1.1 assessment is labor-intensive and variable. This dataset provides the necessary foundation for AI to automate this process, significantly reducing radiologists' workload and improving consistency. For example, fine-tuning nnUNet with this dataset led to substantial Dice score improvements (up to 0.33) for lung tumor segmentation, showcasing the potential for highly accurate automated tumor identification and measurement, which can streamline clinical workflows and improve patient care.
Outcome: Reduced assessment time by 80%
Impact: Enabled more frequent monitoring and personalized treatment adjustments.
Calculate Your Potential ROI
Estimate the potential return on investment for integrating AI-powered medical image analysis into your oncology workflow.
Annual Savings
$0
Hours Reclaimed Annually
0
Your AI Implementation Roadmap
A clear path to integrating advanced AI medical imaging solutions into your operations.
Phase 1: Data Integration & Pre-processing
Securely integrate existing CT datasets and apply standardized pre-processing pipelines, including anonymization and format conversion, ensuring data readiness for AI model training. (4-6 Weeks)
Phase 2: Model Adaptation & Fine-tuning
Leverage foundational models like MedSAM and nnUNet, fine-tuning them with institution-specific data to optimize segmentation accuracy for various lesion types. (8-12 Weeks)
Phase 3: Validation & Clinical Integration
Rigorously validate AI model performance against clinical ground truth and integrate validated models into existing PACS and reporting systems for seamless workflow adoption. (6-10 Weeks)
Ready to Transform Your Oncology Workflow?
Connect with our AI specialists to explore how this dataset and our expertise can drive innovation in your practice.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat