Sawtelle strives to create scalable, easy to use, machine learning & computer vision solutions one API call at time.
For more information, reach out to us here!
Medical image processing pipelines can be at times complex and resource intensive. One basic yet critical step is selecting the scan of interest based on presence of anatomy. When there are thousands and millions of images to analyze, manually selecting the scans of interest becomes very resource intensive. To tackle this bottleneck, we created our flagship Computer Vision API which identifies human body parts in computed tomography (CT) scans - available now in AWS Marketplace.
Our API takes in a CT scan and provides the following anatomical information:
the height of the each detected body parts (head,neck,chest,abdomen,pelvis,lower_limb).
body part completeness for chest, abdomen and pelvis.
The detected body parts can be used as tags to catalog in-house images and used for data retrieval. This anatomical information may also serve to quality check for completeness of body parts including chest, abdomen and pelvis (CAP). Completeness of CAP is particularly useful for oncology related workflows.
A coronal plane from a sample CT scan and its corresponding JSON output from the API is provided below:
![]() head |
![]() neck |
![]() chest |
![]() abdomen |
![]() pelvis |
![]() thigh |
![]() crus |
![]() foot |
{
"body_parts":{
"head":{"height":0.0,"unit":"mm"},
"neck":{"height":36.2,"unit":"mm"},
"chest":{"height":207.9,"unit":"mm"},
"abdomen":{"height":57.2,"unit":"mm"},
"pelvis":{"height":0.0,"unit":"mm"},
"lower_limb":{"height":0.0,"unit":"mm"}
},
"cap_completeness":{"chest":true,"abdomen":false,"pelvis":false}
}
To use this API, subscribe via AWS Marketplace. For more information, check out the API documentation, as well as the technical paper.
Several related academic papers are listed below.
Roth, Holger R., et al. “Anatomy-specific classification of medical images using deep convolutional nets.” 2015 IEEE 12th international symposium on biomedical imaging (ISBI). IEEE, 2015. https://arxiv.org/abs/1504.04003
Sugimori, Hiroyuki. “Classification of computed tomography images in different slice positions using deep learning.” Journal of healthcare engineering 2018 (2018). https://pubmed.ncbi.nlm.nih.gov/30123439
Robert, J. Harris, et al. “High-throughput image labeling and quality control for clinical trials using machine learning.” (2018). https://www.ijclinicaltrials.com/index.php/ijct/article/view/284/165
DISCLAIMER – For Investigational Use Only. The performance characteristics of this product have not been established.