| --- |
| license: cc-by-4.0 |
| tags: |
| - mri-reports |
| - radiology-reports |
| - medical-reports |
| - clinical-reports |
| - medical-nlp |
| - healthcare-ai |
| - biomedical-nlp |
| - report-with-findings |
| - medical-imaging |
| - document-understanding |
| size_categories: |
| - 100K<n<1M |
| --- |
| **This dataset is a large-scale collection of MRI (Magnetic Resonance Imaging) reports with confirmed clinical findings, containing data from 53,852 patients and 29,028,321 medical images, designed to support the development and training of advanced healthcare AI, medical imaging, diagnostic AI, and clinical NLP systems.** |
|
|
| It consists of real-world MRI data where radiological reports indicate the presence of diseases, abnormalities, or pathological conditions. These findings may include tumors, lesions, infections, degenerative changes, or other clinically significant observations. |
| The dataset captures authentic imaging characteristics such as scanner variability, acquisition protocols, and patient positioning, along with detailed clinical narratives. This makes it highly valuable for building accurate, scalable, and production-ready AI systems for medical diagnosis and imaging analysis. |
| Additionally, this dataset can be used in pipelines for Supervised Fine-Tuning (SFT) workflows. |
|
|
| **Dataset Specification** |
|
|
| -Patients: 53,852 |
| -Images: 29,028,321 |
| -Modality: MRI (Magnetic Resonance Imaging) |
| -Type: Medical images with abnormal findings |
| -Data Source: Clinical MRI reports |
| -Body Regions: Brain, Spine, Abdomen, etc. |
| -Data Nature: Real-world clinical data |
| |
| **Key Use Cases** |
|
|
| -Disease detection and classification |
| -Tumor and lesion identification |
| -Medical image segmentation |
| -Clinical decision support systems |
| -Radiology report generation |
| -Multi-modal learning (image + text) |
| -Diagnostic AI model training |
| |
|
|
| **Value of This Dataset** |
|
|
| -Enables supervised training for disease detection |
| -Supports multi-class and multi-label classification |
| -Improves diagnostic accuracy of AI models |
| -Useful for segmentation and localization tasks |
| -Helps build clinically relevant AI systems |
| -Supports real-world healthcare applications |
| |
|
|
| **Quality Analysis** |
|
|
| | Metric | Best Dataset Result | Importance | |
| | --------------------------------- | -------------------------- | --------------------------------------------------------------------- | |
| | **Resolution** | **398×398 avg to 576×576** | Preserves high anatomical and structural detail for accurate analysis | |
| | **SNR (Signal-to-Noise Ratio)** | **25.67** | Indicates strong signal quality with lower noise interference | |
| | **CNR (Contrast-to-Noise Ratio)** | **50.37** | Shows excellent contrast and clear tissue separation | |
| | **Blur Score (Sharpness)** | **363.55** | Reflects extremely sharp and well-defined image quality | |
|
|
|
|
| **Basic JSON Schema** |
| ```json |
| { |
| "patient_id": "large_string", |
| "slice_index": "int64", |
| "modality": "large_string", |
| "findings": "large_string", |
| "image": "Image" |
| } |
| ``` |
| **Data Creation** |
|
|
| Procured through formal agreements and generated in the ordinary course of business. |
|
|
| **Considerations** |
|
|
| This dataset is provided for research and educational purposes only. It contains only sample data. For access to the full dataset and enterprise licensing options, please visit our website [InfoBay.AI](https://infobay.ai/) or contact us directly. |
|
|
| -Ph: (91) 8303174762 |
| -Email: datareq@infobay.ai |
