This module develops the analytical skill of distinguishing AI-assisted tasks from tasks that require clinician judgment or administrative controls, using real medical imaging workflow scenarios as the frame for decision-making.

This module builds the evaluation skills needed to assess AI-generated outputs for quality, traceability, and suitability for deployment in governed healthcare environments, using prepared Microsoft Foundry output scenarios and defined acceptance criteria.

This module covers how to configure and run a supervised learning workflow in Microsoft Foundry (formerly Azure AI Studio) using labeled multimodal datasets—connecting imaging studies, clinical metadata, and patient records—to generate imaging-related prediction outputs.

This module develops the skill of interpreting model behavior by examining how different multimodal input types—imaging-derived variables, report text signals, and structured health data—contribute to prediction outcomes in a bounded healthcare use case.

This module builds the ability to interpret ML model performance outputs using healthcare-relevant metrics—precision, recall, F1, ROC-AUC—and to identify error patterns that affect deployment readiness for imbalanced imaging classification tasks.

This module covers how to select pretrained deep learning architectures and configure training settings in Microsoft Foundry (formerly Azure AI Studio) for bounded medical imaging tasks—without building models from first principles.

This module develops the ability to identify and analyze failure modes in deep learning outputs—specifically missed targets, false-positive signals, and modality-specific weaknesses—using prepared Microsoft Foundry output comparison scenarios.

This module builds the judgment to evaluate transfer-learning options in Microsoft Foundry—determining when a pretrained model can be used as-is versus when additional fine-tuning is needed for a bounded medical imaging workflow.

This module covers how to configure computer vision prompts and model settings in Microsoft Foundry to support bounded detection, classification, and localization tasks for radiology images—within defined clinical-use constraints and without independent clinical interpretation.

This module builds the analytical skill of reviewing computer vision outputs to distinguish between valid image-level findings, regional cues, and artifact-related misreads—a critical competency for healthcare administrators who review or approve AI-assisted radiology workflow outputs.

This module helps learners apply generative AI tools to the multimodal AI and computer vision skills developed in LC 1. It covers how to use AI-assisted prompting to accelerate imaging workflow analysis, how to evaluate generative AI outputs for accuracy and clinical appropriateness, and how to identify where GenAI augmentation adds value versus where human review remains essential.

This project integrates the four core skills—multimodal AI workflow explanation, ML prediction task application, deep learning output analysis, and computer vision output review—in a single realistic healthcare AI evaluation scenario. Learners produce a professional AI workflow evaluation report that mirrors the kind of analysis a healthcare administration professional would prepare before a deployment review meeting.