Medical Coding AI — Part 2: Deep Learning, Transformers & Advanced Architectures

This course builds a complete deep learning system for automated ICD-10-CM coding from clinical notes. Starting from classical ML baselines, students progress through neural word embeddings, CNN and LSTM sequence models, label-wise attention architectures (CAML), imbalanced loss engineering, transformer fundamentals, fine-tuned biomedical language models (ClinicalBERT, PubMedBERT, ClinicalLongformer), and advanced architectures for the long-tail ICD distribution: hierarchical classification, graph neural networks on comorbidity and taxonomy graphs, few-shot learning, and ontology-augmented representations. By the end, students have trained and evaluated a full multi-label ICD coding system with measurable tail-code improvement over the TF-IDF baseline built in Part 1.

  • Level: advanced
  • Duration: 16 weeks
  • 13 sections
  • Price: $19.99

Topics: deep learning, ClinicalBERT, PubMedBERT, ClinicalLongformer, multi-label classification, attention mechanisms, CAML, focal loss, transformer fine-tuning, hierarchical ICD classification, graph neural networks, Spektral, comorbidity graphs, few-shot learning, ontology augmentation, long-tail learning, ICD-10-CM

What's inside

  1. Medical Coding AI — Part 2: Deep Learning, Transformers & Advanced Architectures
  2. Chapter 1: Traditional Machine Learning for Medical Coding
  3. Chapter 2: Evaluation Metrics for Coding Systems
  4. Chapter 3: Neural Networks and Word Embeddings for Clinical Text
  5. Chapter 4: Attention-Based Coding Models
  6. Chapter 5: Imbalanced Learning and Loss Engineering
  7. Chapter 6: Transformer Fundamentals
  8. Chapter 7: Biomedical Language Models
  9. Chapter 8: Long-Document Processing for Clinical Notes
  10. Chapter 9: Hierarchical ICD Classification
  11. Chapter 10: Graph Neural Networks for Medical Coding
  12. Chapter 11: Few-Shot and Zero-Shot Medical Coding
  13. Chapter 12: Ontology-Augmented Learning

Read a free sample chapter