Artificial Intelligence — Deep Learning
Artificial intelligence (AI) — machine and deep learning — involves passing data to a computer algorithm that can build out statistical models that interpret these data. While both traditional machine learning and deep learning models can be complex, deep learning models require significantly more computational resources. The benefit of deep learning models is they achieve humanlike performance for detection and problem classification. For example, these models can detect the presence of tumors in thousands of images. Once trained, a model can generate accurate findings in a fraction of the time required for manual analysis.
The Biomedical Imaging Resource Core can help investigators perform deep or machine learning on large volumes of image data. This allows investigators across disciplines and physical locations to use the virtual platform on campus or remotely with better efficiency and functionality than other offerings in the marketplace.
Investigators can reach out to Biomedical Imaging Resource Core staff with expertise in hardware, algorithms and protocols to help write protocols that are technically and strategically correct and provide consultation on how to collect pilot data for grant applications.
The core's AI service is flexible to provide both novice and experienced users with support tailored to their unique needs.
The Biomedical Imaging Resource Core offers three tiers of AI service:
1. Full-service AI development (high tier)
Target clients: Customers with little to no AI expertise or those wanting a complete solution.
Services: The Biomedical Imaging Resource Core provides end-to-end AI development, including:
- Consulting to identify the customer's problem and how AI can solve it.
- Collecting and processing the client's data.
- Developing, training and testing AI models.
- Deploying and integrating the AI solution into the client's existing systems.
- Offering ongoing support, maintenance and model updates.
Value: Ideal for customers who want an all-in-one solution and lack the in-house resources to manage AI development on their own.
2. Collaborative AI development (mid-tier)
Target clients: Customers with some technical resources or those who want partial assistance in developing AI solutions.
Services: The client collaborates with the Biomedical Imaging Resource Core's AI development team but shares responsibility for certain tasks. For example:
- The client provides the data or helps with initial data preprocessing.
- The Biomedical Imaging Resource Core assists with model selection, training and testing.
- Both parties work together to deploy and integrate the AI model.
Value: Perfect for clients who want some control over the project and may have their own information technology or data science team but still need expert guidance to ensure success.
3. DIY AI development support (low tier)
Target clients: Customers with technical expertise or developers who want access to tools and occasional expert assistance.
Services: The Biomedical Imaging Resource Core provides access to AI development tools, such as high-end graphics processing units and server space for data storage and quick access, that clients can use independently. Also:
- Clients build and train their own AI models.
- The Biomedical Imaging Resource Core offers technical support, troubleshooting and consultation services as needed.
Value: Best suited for customers who want flexibility and control over the AI development process but still value having expert advice available when needed.