Automated Detection and Classification of Tumors from Digitized Histological Tissue Specimens | Rutgers University Innovation Ventures

Automated Detection and Classification of Tumors from Digitized Histological Tissue Specimens

Flowchart illustration of the working of the algorithm


Invention Summary:

Computer-aided diagnosis (CAD) systems can automatically sort digitized medical images which can otherwise be a time-consuming process.

Researchers at Rutgers have developed a CAD method to automatically distinguish between tumor and benign regions in digitized prostate histology. The image analysis is performed in a multi-resolution framework, which is similar to the method in which a pathologist would conduct diagnosis.

The structure of interest is first segmented into smaller images to quickly scan and classify lower-resolution images. This allows for rapid elimination of benign pixels and retention of regions classified as tumors. Additionally, a hierarchical multi-scale classifier is used to enable efficient analysis of large digitized images (1-2 GB). This approach can rapidly detect prostate cancer from digitized data such as MRI images.

Market Applications:

  • Digital pathology
  • Medical imaging
  • Computer-aided diagnosis
  • Image-guided therapy       

Advantages:

  • Classification of lower-resolution images
  • Fully automated and reproducible
  • Accurate distinction between tumor and benign pixels
  • Analysis of large digitized file sizes

Intellectual Property & Development Status:
Issued US patent. US8280132. Available for licensing and/or research collaboration.

 

Publication:

[1]. Doyle, S., Feldman, M., Tomaszewski, J. & Madabhushi, A. (2012). IEEE transactions on bio-med. Engineering. 59, 1205-18. 

Patent Information: