Cancer Registry Coding via Hybrid Neural Symbolic Systems in the Cross-Hospital Setting

Abstract

Cancer registries are critical databases for cancer research whose maintenance requires various types of domain knowledge with labor-intensive data curation. In order to facilitate the curation process with high quality in a timely manner, we developed a hybrid neural symbolic system for cancer registry coding. Unlike previous works which mainly worked on the dataset collected from one hospital or formulated the task as text classification problems, we collaborated with two medical centers in Taiwan to compile a crosshospital corpus and applied neural networks to extract cancer registry variables described in unstructured pathology reports along with expert systems for generating registry coding. We conducted experiments to study the feasibility of the proposed hybrid for the task of cancer registry coding and compare its performance with state-of-the-art non-hybrid approaches. Furthermore, cross-hospital experiments were performed to study the advantages and limitations of transfer learning for processing reports from different sources. The experiment results demonstrated that the proposed hybrid neural symbolic system is a robust approach which works well across hospitals and outperformed classification-based baselines by F-scores of 0.13~0.27. Compared to the baseline models, the F-scores of the proposed approaches are apparently higher when fewer training instances were used. All methods benefited from the transferred parameters learned from the source dataset, but the results suggest that it is a better strategy to transfer the learned knowledge through the concept recognition task followed by the symbolic expert system to address the task of cancer registry coding. INDEX TERMS Electronic medical records, medical expert systems, medical information systems, natural language processingThis work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.

Integrating predictive coding and a user-centric interface for enhanced auditing and quality in cancer registry data

“…For example, the F-scores of “Primary Site” and “Laterality” drop from 0.884 and 0.948 to 0.286 and 0.374, respectively. In contrast to the colorectal cancer registry task investigated in our previous study [12] , the task setting in this study closely mirrors the real hospital setting, wherein all encountered during a patient’s journey are gathered for analysis. Moreover, considering that lung cancer patients may receive diagnoses of multiple primary lung cancers, the inclusion of weighted rules becomes necessary.…”

“…In this study, we enhanced the hybrid neural symbolic system originally introduced in our prior research [12] . This enhancement involved the incorporation of weighted rules specifically designed for determining the target coding items from the various reports found within a patient’s medical history.…”

“…To expedite our development process for the lung cancer registry coding task, we harnessed the power of transfer learning. Specifically, we leveraged our established colorectal cancer registry concept recognition model [12] as a pre-trained model. To adapt it for the specific requirements of this study, we modified the model by increasing the number of output nodes in its final multilayer perceptron layer to align with the number of target concept types defined for lung cancer registry tasks.…”

“…In particular, they highlighted the pre-training process based on TF-IDF-weighted word embeddings can improve the performance in the primary site classification task. Another study presented by Dai, Yang [12] proposed a hybrid neural symbolic method for cancer registry coding. Their research primarily focused on the challenge of transfer learning across different hospitals.…”

Epoch in a neural network for brain stroke

“…Ischemic often occurs with experience the clot that can cover Genre blood in the brain. Need conducted operation to remove lumps them as well as need intensive treatment to better again [5]. While hemorrhagic stroke occurs if there is bleeding in the brain [6].…”