Iterative Annotation to Ease Neural Network Training: Specialized Machine Learning in Pathology


Expert-trainable automated medical image segmentation and annotation AI with user-friendly interface




Automated diagnosis in medical imaging has been hindered by a lack of well-annotated images. Neural network-facilitated annotation is a potential solution to this problem, but ‘supervised’ machine learning using images annotated by human experts is often costly and time consuming, while small dataset sizes in medical imaging have meant that fully automated ‘unsupervised’ AI segmentation of image data has thus far been impractical. 


Technology Overview:


This technology addresses the above problem by combining unsupervised and supervised learning with a user-friendly annotation interface. The software is used to annotate a small training dataset to partially train an image segmentation and annotation AI. The annotations performed by the AI are displayed on the image itself and can be easily corrected by the user via ‘clicking and dragging’. After a few rounds of iterative training, a highly accurate trained AI can be produced and deployed.


The software has broad applications and has already been applied to kidney histology slides and MRI images of human prostate.





•       Semi-automated image annotation and segmentation

•       Intuitive user interface

•       Rapid AI development and deployment




•       Histology / Pathology

•       Medical Imaging

•       Microscopy


Intellectual Property Summary:


Copyright & Know-how


Stage of Development:


•       Technology validated in lab, TRL 4


Licensing Status:


Available for License


Additional Information:


Patent Information:
For Information, Contact:
Michael Fowler
Commercialization Manager
University at Buffalo
© 2020. All Rights Reserved. Powered by Inteum