Malignancy of breast lesions on diffusion weighted images
| Purpose | Detect breast lesions and determine malignancy of breast lesions on diffusion weighted images using ADC values |
| Tag(s) | |
| Panel | Breast Imaging |
| Define-AI ID | 19060010 |
| Originator | Vandana Dialani |
| Lead | Vandana Dialani |
| Panel Chair | Elizabeth Burnside |
| Panel Reviewers | Breast Imaging |
| License | |
| Status | |
| RadElement Set | RDES132 |
Clinical Implementation
Value Proposition
Breast MRI is commonly used to detect and diagnose breast cancer, though there are some limitations with T1/T2 weighted images to describe lesions. Use of diffusion in triaging biopsies would be more robust with AI and could help reduce biopsies and increase specificity of breast MRI, thus reducing overdiagnosis.
In addition the current application of DWI is in staging, and evaluation of treatment response. The lower the ADC values the higher the grade of IDC. AI could bring uniformity in obtaining to ADC values and improve the specificity of DWI — empowering its use as a non contrast MRI noninferior to DCE MRI. Eventually DWI could be widely utilized as a non contrast screening modality.
Narrative(s)
Patient has a breast mass which is indeterminate on MRI( Birads 4a or 4b) and biopsy recommended. With this technique clinician could increase specificity of masses which undergo biopsy and potentially save some biopsies.
Workflow Description
The image is obtained from modality and sent to PACS and the AI engine. The image is analyzed by the engine. The system then detects breast lesion and estimates malignancy. An alert message is sent to PACS from the engine with the information, identification, and graphic highlighting the malignant components.
Diagnostic MRI
1) Breast mass indeterminate on MRI( Birads 4a, 4b, 4c) and biopsy recommended.
2) Patient on neoadjuvant chemotherapy, adc obtained at different points of treatment for assessing treatment response
Screening MRI
1) identify breast lesion/ breast cancer
2) obtain adc maps/values
Considerations for Dataset Development
| Age | >30 years |
| malignant lesion | invasive ductal carcinomas, Ductal carcinoma in situ, invasive lobular carcinoma, lobular carcinoma in situ, angiosarcomas, inflammatory breast cancer, paget disease of the breast, phyllodes tumors |
| Lesion size | varied |
Technical Specifications
Inputs
DICOM Study
| Procedure | Diffusion weighted MRI |
| Views | axial, coronal, sagittal |
| Data Type | DICOM |
| Modality | MR |
| Body Region | Chest |
| Anatomic Focus | Breast |
| Pharmaceutical | N/A |
| Scenario | Diffusion weighted MRI |
Primary Outputs
Malignant breast lesion detection
| RadElement ID | RDE872 |
| Definition | Detection of malignant breast lesion |
| Data Type | Categorical |
| Value Set |
|
| Units | N/A |
ADC values on ROI
| RadElement ID | RDE873 |
| Definition | ADC value on ROI |
| Data Type | Numeric |
| Value Set | N/A |
| Units | mm2/s |
Future Development Ideas
Segment the malignant component of a lesion
Related Datasets
The Cancer Imaging Archive (TCIA)
| |
Know of more related datasets? Please let us know