Skip to content

Quiz: AI and Machine Learning Integration

Test your understanding of artificial intelligence and machine learning integration with graph databases for advanced healthcare applications.

1. What is the definition of Artificial Intelligence in healthcare contexts?

  1. Computer systems capable of performing tasks that traditionally require human intelligence
  2. A database technology optimized for storing medical images
  3. A programming language designed specifically for healthcare applications
  4. A regulatory framework governing medical device software
Show Answer

The correct answer is A. Artificial Intelligence (AI) refers to computer systems capable of performing tasks that traditionally require human intelligence, including reasoning, learning, perception, and language understanding. In healthcare contexts, AI systems analyze clinical data, identify patterns, recommend treatments, and support diagnostic processes. Options B, C, and D describe other technologies or frameworks but are not definitions of artificial intelligence.

Concept Tested: Artificial Intelligence

See: AI and ML Fundamentals

2. What does RAG stand for in the context of healthcare AI systems?

  1. Random Access Generation
  2. Regional Analytics Gateway
  3. Retrieval-Augmented Generation
  4. Relational Algorithm Graph
Show Answer

The correct answer is C. RAG stands for Retrieval-Augmented Generation, an architecture that combines retrieval of relevant factual information with the generative capabilities of large language models. This hybrid approach provides LLMs with current, domain-specific knowledge at query time, dramatically reducing hallucination risk while enabling personalized responses based on specific patient data. Options A, B, and D are not correct acronyms used in healthcare AI.

Concept Tested: RAG Architecture

See: RAG Architecture: Combining Retrieval and Generation

3. Which best describes the hierarchical relationship between Artificial Intelligence, Machine Learning, and Deep Learning?

  1. They are three completely independent technologies
  2. AI encompasses ML, which encompasses Deep Learning as nested subsets
  3. Deep Learning is the broadest category, containing ML and AI
  4. Machine Learning and Deep Learning are separate from AI
Show Answer

The correct answer is B. Artificial Intelligence encompasses all intelligent systems, Machine Learning is a subset of AI that uses data-driven learning, and Deep Learning is a subset of ML that uses neural networks with multiple layers. This hierarchical relationship can be visualized as concentric circles with AI as the outer circle, ML in the middle, and Deep Learning at the center. Options A, C, and D incorrectly represent the relationship between these technologies.

Concept Tested: Machine Learning

See: Artificial Intelligence and Machine Learning Fundamentals

4. Why are graph databases and large language models considered complementary technologies in healthcare AI?

  1. They both use the same underlying neural network architecture
  2. They both require the same amount of training data
  3. They both solve identical problems in different ways
  4. Graphs excel at structured reasoning while LLMs excel at natural language understanding
Show Answer

The correct answer is D. Graph databases and LLMs are fundamentally complementary because graphs excel at explicit, verifiable relationships and logical reasoning, while LLMs excel at handling ambiguous natural language, recognizing patterns in unstructured text, and generating human-friendly explanations. This complementary nature allows integrated systems to leverage the precision of graph queries for safety checks (contraindications, allergies) while using LLM capabilities for understanding questions and generating contextualized responses. Options A, B, and C are incorrect as these technologies have different architectures, data requirements, and complementary (not identical) capabilities.

Concept Tested: Graph And LLM Integration

See: Graph and LLM Integration: Complementary Strengths

5. A hospital wants to build a system that answers clinician questions about treatment protocols using the latest research. Which approach would be most effective?

  1. A standalone relational database with keyword search
  2. A RAG system combining vector store retrieval with LLM generation
  3. A rule-based expert system with predefined responses
  4. A standalone large language model trained on general medical knowledge
Show Answer

The correct answer is B. A RAG (Retrieval-Augmented Generation) system would be most effective because it combines retrieval of the latest research from a vector store with LLM generation capabilities to provide accurate, current, explainable responses with source citations. This addresses the key limitations of standalone LLMs (knowledge cutoff, hallucination risk) while enabling natural language interaction. Option A lacks semantic understanding, Option C cannot handle novel questions or updated research, and Option D suffers from knowledge cutoff and hallucination risks without grounding in current research.

Concept Tested: RAG Architecture

See: RAG Architecture: Combining Retrieval and Generation

6. Given a clinical query "patient experiencing chest pain and shortness of breath," how do vector embeddings improve search results compared to keyword matching?

  1. Vector embeddings find semantically similar content even with different terminology like "cardiac arrest" or "myocardial infarction"
  2. Vector embeddings only match exact character strings for precise results
  3. Vector embeddings eliminate the need for any medical knowledge in the system
  4. Vector embeddings work faster than keyword search for all query types
Show Answer

The correct answer is A. Vector embeddings transform medical concepts into numerical representations that capture semantic meaning in high-dimensional space, placing semantically similar concepts close together even when they use different terminology. Unlike keyword matching that treats "cardiac arrest" and "heart stopped" as completely different, embeddings recognize them as highly related concepts and return relevant results regardless of exact word matching. Option B describes keyword matching, not embeddings. Option C is incorrect as embeddings are learned from medical knowledge. Option D is incorrect as performance depends on implementation, not inherent to the approach.

Concept Tested: Vector Embedding

See: Vector Embeddings and Semantic Representation

7. A physician orders a new medication for a patient with multiple chronic conditions. Which component of a graph-based clinical decision support system would identify potential drug-disease contraindications?

  1. The LLM interface that generates natural language explanations
  2. The ML risk models that predict readmission probability
  3. The rules engine that traverses patient and knowledge graphs
  4. The vector store containing research literature embeddings
Show Answer

The correct answer is C. The rules engine evaluates conditions by traversing both the patient graph (containing the patient's current diagnoses, medications, allergies, and lab results) and the clinical knowledge graph (containing evidence-based guidelines, medication information, and contraindications). This traversal identifies relationships such as drug-disease contraindications by querying paths like medication CONTRAINDICATED_IN specific patient conditions. Options A, B, and D are components of the system but do not perform the contraindication checking through graph traversal.

Concept Tested: Clinical Decision Support

See: Clinical Decision Support Systems

8. What is the primary limitation of standalone large language models that makes them risky for clinical decision-making?

  1. They cannot process natural language queries effectively
  2. They require too much computational power for healthcare settings
  3. They lack integration with existing hospital IT systems
  4. They may generate plausible-sounding but factually incorrect information (hallucination)
Show Answer

The correct answer is D. The most critical limitation of standalone LLMs for clinical use is their tendency to hallucinate—generating plausible-sounding but factually incorrect information. This occurs because LLMs are trained to produce fluent text based on patterns, not verified facts. In clinical contexts where accuracy is critical for patient safety, this hallucination risk is dangerous. Additional limitations include knowledge cutoff, lack of patient-specific context, and explainability challenges. Options A is incorrect as LLMs excel at natural language, B is a practical concern but not the primary safety risk, and C is an integration challenge, not an inherent limitation.

Concept Tested: Large Language Model

See: Large Language Models in Healthcare

9. When would Graph Neural Networks be most appropriate for healthcare predictive analytics compared to traditional machine learning approaches?

  1. When predicting outcomes based solely on individual patient demographics
  2. When the dataset contains only structured tabular data with no relationships
  3. When computational resources are extremely limited
  4. When patient outcomes depend on relational context like care coordination and social networks
Show Answer

The correct answer is D. Graph Neural Networks (GNNs) are most appropriate when patient outcomes depend on relational context that traditional ML approaches would miss. GNNs learn to aggregate information from a node's neighborhood, capturing how outcomes depend not just on patient attributes but on the broader clinical context including provider connections, social support networks, care coordination patterns, and comorbidity interactions. For example, readmission risk may depend on whether a patient has strong family connections (social network edges) or fragmented care across unconnected providers. Options A and B describe scenarios better suited to traditional ML, and Option C is incorrect as GNNs typically require more computational resources than simpler models.

Concept Tested: Graph Neural Network

See: Predictive Analytics for Proactive Care

10. A healthcare organization must choose between three approaches for clinical question answering: (1) standalone graph database, (2) standalone LLM, or (3) integrated graph+LLM system. Which factors should most heavily influence this decision?

  1. Only the initial implementation cost
  2. The balance between factual accuracy, natural language capability, and explainability requirements
  3. Which vendor offers the longest support contract
  4. The programming language preferences of the development team
Show Answer

The correct answer is B. The decision should be based on balancing critical clinical and technical requirements: factual accuracy (graphs excel, LLMs risk hallucination), natural language understanding (LLMs excel, graphs require structured queries), and explainability (graphs provide clear paths, LLMs are harder to explain). An integrated system combines the strengths of both—using graphs for factual grounding and safety checks while leveraging LLMs for natural language interaction and explanation generation. This achieves high scores across all three dimensions. While cost (A), vendor support (C), and technical preferences (D) are considerations, they should not override the fundamental capability requirements for safe, effective clinical decision support.

Concept Tested: Graph And LLM Integration

See: Graph and LLM Integration: Complementary Strengths

Quiz Complete

Concepts Covered: - Artificial Intelligence - Machine Learning - Large Language Model - Vector Embedding - RAG Architecture - Graph And LLM Integration - Clinical Decision Support - Graph Neural Network

Bloom's Taxonomy Distribution: - Remember: 20% - Understand: 20% - Apply: 30% - Analyze: 20% - Evaluate: 10%

For more information on these concepts, review Chapter 10: AI and Machine Learning Integration.