
Dr. Balis is a Board-certified pathologist with subspecialty boards in Clinical Informatics, who has maintained longstanding interest in the intersection of engineering, computational approaches and the practice of medicine. He is a professor of Pathology at the University of Michigan and currently serves as the director of the Division of Pathology Informatics, in the Department of Pathology. This division is noteworthy for being one of the few such academic information technology divisions operating in support of pathology while being housed wholly within a pathology department itself.
He has active, NIH R01-supported research initiatives in several areas of pathology and medical informatics, including machine learning and use of encoded data, image-based analytics, machine vision tools for histopathology, image-based search algorithms and federated enterprise data architectures with all of these areas serving as rich training substrate for a growing and thriving pathology fellowship program. He has delivered over 200 invited presentations, nationally and internationally, on various topics related to pathology informatics, image analysis and data analytics.
TITLE: Empowering Pathologists to Directly Interrogate Digital Histology Subject Matter with Open Source – Introducing the VIPER Studio Histology Machine Learning Platform as an Exemplar”
Abstract:
Contemporary efforts to apply machine vision and convolutional Neural Network (CNN) based analyses to histopathological subject matter has already demonstrated significant utility for both general image classification tasks, as well as for implementation of unsupervised partitioning of datasets into multiple appropriate diagnostic subclasses. However, current methodologies are limited in that the required tools, methods and computational pipelines necessarily to realize these tasks impart the need for significant technical expertise in critical steps, such as: ground truth generation, feature vector design, and overall pipeline optimization. These requirements have the potential to impose barriers to pathologists being an integral component of the overall discovery process. In this presentation, Dr. Balis will introduce the availability of a new, open-source tool, VIPER Studio, which is specifically designed to allow for pathologists to generate feature segmentation libraries, ground truth map, and complete CNN pipelines, all without the associated requirement of a technical team to be available. Moreover, by use of hand-crafter features, it can be seen that the typically required 100s to 1000s of training set images can be reduced to far smaller numbers, without a significant loss in ROC performance.
The multi-stage approach intrinsic to VIPER Studio has been applied to a number of histology classification use-cases, with preliminary results demonstrating that in many cases, the requirement for having hundreds or thousands of images can be significantly reduced to requiring, in most instances, only a few representational images. Live demonstrations of the software will be employed, depicting a number of salient use cases.