![]() The SimpleMind framework provides both completed image analysis solutions for several public datasets,Īs well as a software development toolkit (SDK) to build a custom model. Many medical imaging applications, and we believe that there is strong potential utilityįor other research and commercial groups. ![]() In development at our research center since 2016, the SM framework has improved the performance of Making it easier for a data scientist to teach, train, and optimize the AI, andīetter segmentation accuracy and reliability, with common sense reasoning to avoid obvious mistakes. The increased level of intelligence brings two major benefits: It provides end-to-end automatic parameter tuning (APT) to intelligently explore combinations of agent parameters, enabling a more extensive, unbiased search than manual tuning by a human data scientist and resulting in a more optional solution. Using this model, SM can automatically chain together a series of general-purpose agents for image pre/post processing DNNs, and machine reasoning. A high level model represents knowledge usually incorporated ad hoc by the data scientist, including descriptions of scene content and concepts and meta knowledge about learning and image processing. The SM framework brings transparency and automation to the tasks of the data scientist. Crucially, reasoning can be used to avoid common sense (“dumb”) mistakes that violate basic high-level concepts (e.g., anatomical constraints). It can provide a layer of machine reasoning atop DNNs, applying knowledge where representative training data may be limited and enabling conceptual decision making (e.g., whether a device is properly positioned relative to an anatomical landmark). Cognitive AI is broadly defined as enabling human-level reasoning and intelligence. These shortfalls impact the performance of AI systems and leaves them vulnerable to errors that are obvious to a human, resulting in a loss of trust and limiting real-world clinical application and adoption.Ĭognitive AI can address the limitations of Narrow AI by melding high-level conceptual knowledge with pattern recognition. The result is application-specific Narrow AI, that is typically suboptimal in terms of parameter search and limited as to the level of knowledge and reasoning applied. This knowledge is coded ad hoc in scripts, with limited application of common sense reasoning and limited hand tuning of parameters (both in learning and pre/post processing). The current applications are in medical image segmentation and analysis.ĭeep learning segmentation algorithms are data driven, but rely on human (data scientist) knowledge for hand tuning of deep neural network (DNN) architectures and learning hyper parameters and applying knowledge ad hoc in heuristic pre/post processing algorithms. ![]() It makes it easier for a data scientist to develop knowledge-based AI applications and can achieve better segmentation accuracy and reliability in clinical practice. ![]() The framework can be used to add reasoning to pre-existing DNNs and/or to train DNNs with pre/post processing and end-to-end parameter optimization. SimpleMind (SM) is an open source software framework that supports deep neural networks (DNNs) with higher level machine reasoning and automatic parameter tuning. A virtual data scientist built using Cognitive AI Cognitive AI Framework
0 Comments
Leave a Reply. |