We are looking for medical imaging experts to acquire radiology datasets from local hospitals, imaging centers, and radiology practices. Specifically, we want to acquire frontal chest X-rays (XR), with computed tomography (CT) scans performed within 6 months of each other, along with indications that are natural language processed from corresponding radiology reports. These images will be used to train computer vision algorithms that will assist radiologists in finding early evidence of lung cancer.
You can help
Lung cancer is a worldwide problem, and as a result we are curating an international dataset to further refine and validate computer vision algorithms that will help reduce overlooked lung nodules. By doing so, you will help us further the research in lung cancer, help improve the diagnostic accuracy of algorithms, and ultimately help us on our mission to save lives with data science.
The project is broken down into five phases:
1. Phase I: Data Availability and Qualification
o We need to quantify and qualify the data that you can acquire for this project. To do so, we will want responses to the following questions:
-A list of local hospitals that you can enlist in participating in this project.
-The number of digitally acquired chest XR's that can be acquired. These are typically stored in a picture archiving and communication system (PACS).
-The number of chest CT's that can be acquired.
-If corresponding radiology reports for the CT's and XR's can be natural language processed (NLP). This is typically stored in a Radiology Information System (RIS). We will provide the NLP utility.
-If pathology reports for the patients with CT's and XR's can be natural language processed.
-Patient identifier code (or the like, commonly called medical record number or MRN) to link XR, CT, RIS / radiology, and pathology reports by patient.
o Deliverable: Responses to the above questions.
2. Phase II: Data Validation
o Assuming the phase I deliverable meets our needs, we will proceed with a small sample of data for validation.
-As appropriate and needed, attain authorizations from hospitals, radiology practices or imaging centers to share data for research purposes.
-Send a small sample of 10 de-identified CT confirmed XR's via DVD or secure file transfer (SFTP).
-Our radiologist will review for acceptability for our research purposes.
o Deliverable: 10 Sample CT's and XR's (DICOM format)
3. Phase III: Data Acquisition
o Assuming the sample data passes validation in Phase II, we will proceed with the full data acquisition:
-Pull full historical Chest XR data out of the PACS.
-Pull full historical CT data (related to XR's) out of the PACS.
-Pull full historical RIS / radiology report data (related to XR's).
-Pull full historical pathology reports (related to XR's).
4. Phase IV: Data De-Identification
o To comply with standard privacy practices, the full historical dataset will need to be de-identified before sending to us. This will entail:
-Run DICOM information scrubber to remove personal health information on the CT's / XR's. We can provide utilities to do so.
-Run Natural Language Processing (NLP) utility that we provided on radiology and pathology reports. This will determine whether the report indicated potential lung cancer and will remove any personal health information (PHI).
o Deliverable: Fully de-identified and NLP-indexed DICOM data, sent to us via secure file transfer (SFTP) or physical media (such as an external hard drive). We will provide the external hard drive and shipping labels if needed.
5. Phase V: Final Data Verification
o Once the data has sent, we will perform a final verification of the data to ensure:
-DICOM images are readable.
-All files are cross reference-able using MRN or patient identifier.
Skills required
- Knowledge of DICOM medical image format
- Data extraction
- De-Identification of data
- Previous experience working with local PACS systems to retrieve medical images
- Knowledge and adherence to local health information privacy laws
- Local language & English
Contract Terms
Qualified individuals please submit your fixed fee proposals for each phase I-IV. Compensation will be provided at the end of Phase I, II, and V. Please include local institutions that you intend to engage for Phase I along with any answers to phase I questions that are known .
The acquisition and transfer of these images must comply with local health information privacy laws. I will share service terms agreement.
Keywords:
medical, medicine, public health, health technology, DICOM, communication, technical reports, healthcare, research, radiology, x-ray, ct, pathology, lung cancer, cancer
About the recuiterMember since Mar 14, 2020 Pradeep Pandey
from Yucatan, Mexico