Automatic ICD-10-CM coding via Lambda-Scaled attention based deep learning model.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Health Data Management

Background:

• The International Classification of Diseases (ICD)-10-CM is a global standard for healthcare administration, crucial for accurate disease classification.
• Despite widespread adoption, public hospitals in Pakistan have not yet implemented automated ICD-10-CM coding.
• The complexity of medical data necessitates advanced methods for accurate diagnostic classification.

Purpose of the Study:

• To develop and evaluate a novel deep learning model for automated ICD-10-CM coding.
• To address the attention score vanishing problem in existing deep recurrent convolutional neural network (DRCNN) models for this task.
• To implement and deploy the proposed model for practical application in healthcare settings.

Main Methods:

• A private Clinical Pool of Liver Transplant (CPLT) database was utilized for implementing ICD-10-CM coding.
• A novel deep learning model, Deep Recurrent-Convolution Neural Network with a lambda-scaled Attention module (DRCNN-ATT), was proposed.
• The DRCNN-ATT model integrates bi-directional long short-term memory (bi-LSTM), multi-scale convolutional neural network (MS-CNN), and a lambda-scaled attention module.

Main Results:

• The proposed lambda-scaled attention module effectively resolved the attention score vanishing problem observed in standard DRCNN models.
• The DRCNN-ATT model achieved high performance, with micro F1 scores of 0.862 on the CPLT dataset and 0.705 on the MIMIC III top 50 dataset.
• The model was successfully deployed using ngrok and Flask APIs for real-time automatic ICD-10-CM coding.

Conclusions:

• The DRCNN-ATT model represents a significant advancement in automated ICD-10-CM coding, offering improved accuracy and efficiency.
• The successful deployment highlights the model's potential for practical integration into hospital information systems.
• This research contributes to modernizing healthcare administration through advanced AI techniques.