Hierarchical label-wise attention transformer model for explainable ICD coding

Article Computer Science, Theory & Methods

Multi-channel, convolutional attention based neural model for automated diagnostic coding of unstructured patient discharge summaries

Veena Mayya et al.

Summary: Effective coding of patient records in hospitals is crucial for epidemiology, billing, and managing insurance claims. Manual coding is prone to errors and time-consuming, prompting the development of diagnostic coding systems based on patients' Electronic Medical Records. Developing nations with low digitization rates require systems that leverage unstructured data sources.

FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE (2021)

Add to Collection

Article Computer Science, Interdisciplinary Applications

Explainable automated coding of clinical notes using hierarchical label-wise attention networks and label embedding initialisation

Hang Dong et al.

Summary: The study introduces a new Hierarchical Labelwise Attention Network (HLAN) for model explanation and improving the efficiency of medical coding by highlighting key words and sentences to provide more meaningful and comprehensive model interpretation. Additionally, by introducing a label embedding initialization approach, the performance of automated coding can be consistently boosted, capturing the correlations among labels.

JOURNAL OF BIOMEDICAL INFORMATICS (2021)

Add to Collection

Article Computer Science, Information Systems

Limitations of Transformers on Clinical Text Classification

Shang Gao et al.

Summary: This work introduces four methods to scale BERT for document classification on long clinical texts, finding that BERT often cannot beat simpler baseline models in this scenario. Analysis shows that BERT's pretraining and WordPiece tokenization may be limiting its performance in such tasks.

IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS (2021)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence

TransICD: Transformer Based Code-Wise Attention Model for Explainable ICD Coding

Biplob Biswas et al.

Summary: ICD coding procedure is crucial for the medical billing system, but manual assignment of ICD codes may result in errors, making automation an important task. A transformer-based architecture with code-wise attention mechanism is proposed to automate the process, outperforming other baselines and providing more insights to support clinical decisions.

ARTIFICIAL INTELLIGENCE IN MEDICINE (AIME 2021) (2021)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence

Embed Wisely: An Ensemble Approach to Predict ICD Coding

Pavithra Rajendran et al.

Summary: This research enhances the performance of ICD code prediction by combining multiple sets of pretrained word embeddings using post-processing and meta-embeddings techniques, and infusing information from biomedical articles to improve deep learning architectures. It demonstrates the efficacy of this approach for a multi-modal setting, using both unstructured and structured information.

MACHINE LEARNING AND PRINCIPLES AND PRACTICE OF KNOWLEDGE DISCOVERY IN DATABASES, PT II (2021)

Add to Collection

Article Biochemical Research Methods