Multi-label Few-shot ICD Coding as Autoregressive Generation with Prompt.

Automatic International Classification of Diseases (ICD) coding aims to assign multiple ICD codes to a medical note with an average of 3,000+ tokens. This task is challenging due to the high-dimensional space of multi-label assignment (155,000+ ICD code candidates) and the long-tail challenge - Many ICD codes are infrequently assigned yet infrequent ICD codes are important clinically. This study addresses the long-tail challenge by transforming this multi-label classification task into an autoregressive generation task. Specifically, we first introduce a novel pretraining objective to generate free text diagnoses and procedures using the SOAP structure, the medical logic physicians use for note documentation. Second, instead of directly predicting the high dimensional space of ICD codes, our model generates the lower dimension of text descriptions, which then infers ICD codes. Third, we designed a novel prompt template for multi-label classification. We evaluate our Generation with Prompt (GPsoap) model with the benchmark of all code assignment (MIMIC-III-full) and few shot ICD code assignment evaluation benchmark (MIMIC-III-few). Experiments on MIMIC-III-few show that our model performs with a marco F130.2, which substantially outperforms the previous MIMIC-III-full SOTA model (marco F1 4.3) and the model specifically designed for few/zero shot setting (marco F1 18.7). Finally, we design a novel ensemble learner, a cross-attention reranker with prompts, to integrate previous SOTA and our best few-shot coding predictions. Experiments on MIMIC-III-full show that our ensemble learner substantially improves both macro and micro F1, from 10.4 to 14.6 and from 58.2 to 59.1, respectively.

MedJEx: A Medical Jargon Extraction Model with Wiki's Hyperlink Span and Contextualized Masked Language Model Score.

This paper proposes a new natural language processing (NLP) application for identifying medical jargon terms potentially difficult for patients to comprehend from electronic health record (EHR) notes. We first present a novel and publicly available dataset with expert-annotated medical jargon terms from 18K+ EHR note sentences (MedJ). Then, we introduce a novel medical jargon extraction (MedJEx) model which has been shown to outperform existing state-of-the-art NLP models. First, MedJEx improved the overall performance when it was trained on an auxiliary Wikipedia hyperlink span dataset, where hyperlink spans provide additional Wikipedia articles to explain the spans (or terms), and then fine-tuned on the annotated MedJ data. Secondly, we found that a contextualized masked language model score was beneficial for detecting domain-specific unfamiliar jargon terms. Moreover, our results show that training on the auxiliary Wikipedia hyperlink span datasets improved six out of eight biomedical named entity recognition benchmark datasets. Both MedJ and MedJEx are publicly available.

Gut Microbiome and Metabolome Profiles Associated with High-Fat Diet in Mice.

Obesity can be caused by microbes producing metabolites; it is thus important to determine the correlation between gut microbes and metabolites. This study aimed to identify gut microbiota-metabolomic signatures that change with a high-fat diet and understand the underlying mechanisms. To investigate the profiles of the gut microbiota and metabolites that changed after a 60% fat diet for 8 weeks, 16S rRNA gene amplicon sequencing and gas chromatography-mass spectrometry (GC-MS)-based metabolomic analyses were performed. Mice belonging to the HFD group showed a significant decrease in the relative abundance of Bacteroidetes but an increase in the relative abundance of Firmicutes compared to the control group. The relative abundance of Firmicutes, such as Lactococcus, Blautia, Lachnoclostridium, Oscillibacter, Ruminiclostridium, Harryflintia, Lactobacillus, Oscillospira, and Erysipelatoclostridium, was significantly higher in the HFD group than in the control group. The increased relative abundance of Firmicutes in the HFD group was positively correlated with fecal ribose, hypoxanthine, fructose, glycolic acid, ornithine, serum inositol, tyrosine, and glycine. Metabolic pathways affected by a high fat diet on serum were involved in aminoacyl-tRNA biosynthesis, glycine, serine and threonine metabolism, cysteine and methionine metabolism, glyoxylate and dicarboxylate metabolism, and phenylalanine, tyrosine, and trypto-phan biosynthesis. This study provides insight into the dysbiosis of gut microbiota and metabolites altered by HFD and may help to understand the mechanisms underlying obesity mediated by gut microbiota.

Identification of Salivary Microorganisms and Metabolites Associated with Halitosis.

Halitosis is mainly caused by the action of oral microbes. The purpose of this study was to investigate the differences in salivary microbes and metabolites between subjects with and without halitosis. Of the 52 participants, 22 were classified into the halitosis group by the volatile sulfur compound analysis on breath samples. The 16S rRNA gene amplicon sequencing and metabolomics approaches were used to investigate the difference in microbes and metabolites in saliva of the control and halitosis groups. The profiles of microbiota and metabolites were relatively different between the halitosis and control groups. The relative abundances of Prevotella, Alloprevotella, and Megasphaera were significantly higher in the halitosis group. In contrast, the relative abundances of Streptococcus, Rothia, and Haemophilus were considerably higher in the control group. The levels of 5-aminovaleric acid and n-acetylornithine were significantly higher in the halitosis group. The correlation between identified metabolites and microbiota reveals that Alloprevotella and Prevotella might be related to the cadaverine and putrescine pathways that cause halitosis. This study could provide insight into the mechanisms of halitosis.