Extracting Biomedical Entity Relations using Biological Interaction Knowledge

Interaction between visual impairment and subjective cognitive complaints on physical activity impairment in U.S. older adults: NHANES 2005-2008

BACKGROUND/AIM

To investigate the independent relationships of visual impairment (VI) and Subjective cognitive complaints (SCC) with physical function impairment (PFI) and the interaction effect between VI and SCC on PFI in American older adults.

METHODS

The data of this cross-sectional study was obtained from the 2005-2008 National Health and Examination Survey (NHANES) conducted in the United States. The VI criterion included both subjective self-reported eyesight conditions and objective visual acuity test results. The self-reported questionnaires were utilized to determine PFI and SCC. According to the survey design of NHANS, original data were weighted to produce nationally representative estimates. Both the unweighted original data and weighted estimates underwent analysis. Crude and adjusted logistic models were employed to assess the pairwise associations among VI, SCC, and PFI. To assess the interactive effect, measures such as the relative excess risk due to interaction (RERI), attributable proportion due to interaction (AP), and synergy index (S) were calculated.

RESULTS

A total of 2,710 subjects (weighted n = 38,966,687) aged 60 years or older were included. Compared with subjects without subjective visual impairment (SVI), those with SVI had a significant positive association with PFI [weighted OR (95%CI): 3.11 (2.25, 4.31)]. After multi-variable adjusting, the relationship remained significant [weighted OR (95%CI): 1.90 (1.32, 2.72)]. Similarly, those with objective visual impairment (OVI) were positively associated with the risk of PFI in the crude model [weighted OR (95%CI): 2.35 (1.53, 3.61)] and adjusted model [weighted OR (95%CI): 1.84 (1.07, 3.17)]. Moreover, we found the association of SCC with an increased risk of FPI [crude weighted OR (95%CI): 5.02 (3.40, 7.40); adjusted weighted OR (95%CI): 3.29 (2.01, 5.38)]. Ultimately, the additive interaction showed there was a significant positive interaction term between SVI and SCC on PFI, while OVI and SCC did not.

CONCLUSION

Both VI and SCC were significantly associated with PFI in elder adults. Besides, there was a significant synergistic interaction between SVI and SCC on PFI, which indicated the improvement of SVI and SCC may be beneficial for the prevention of PFI. For the elderly, especially those with multiple disabilities, comprehensive and targeted approaches are imperative to foster their overall well-being and health.

Few-shot learning for medical text: A review of advances, trends, and opportunities

BACKGROUND

Few-shot learning (FSL) is a class of machine learning methods that require small numbers of labeled instances for training. With many medical topics having limited annotated text-based data in practical settings, FSL-based natural language processing (NLP) holds substantial promise. We aimed to conduct a review to explore the current state of FSL methods for medical NLP.

METHODS

We searched for articles published between January 2016 and October 2022 using PubMed/Medline, Embase, ACL Anthology, and IEEE Xplore Digital Library. We also searched the preprint servers (e.g., arXiv, medRxiv, and bioRxiv) via Google Scholar to identify the latest relevant methods. We included all articles that involved FSL and any form of medical text. We abstracted articles based on the data source, target task, training set size, primary method(s)/approach(es), and evaluation metric(s).

RESULTS

Fifty-one articles met our inclusion criteria-all published after 2018, and most since 2020 (42/51; 82%). Concept extraction/named entity recognition was the most frequently addressed task (21/51; 41%), followed by text classification (16/51; 31%). Thirty-two (61%) articles reconstructed existing datasets to fit few-shot scenarios, and MIMIC-III was the most frequently used dataset (10/51; 20%). 77% of the articles attempted to incorporate prior knowledge to augment the small datasets available for training. Common methods included FSL with attention mechanisms (20/51; 39%), prototypical networks (11/51; 22%), meta-learning (7/51; 14%), and prompt-based learning methods, the latter being particularly popular since 2021. Benchmarking experiments demonstrated relative underperformance of FSL methods on biomedical NLP tasks.

CONCLUSION

Despite the potential for FSL in biomedical NLP, progress has been limited. This may be attributed to the rarity of specialized data, lack of standardized evaluation criteria, and the underperformance of FSL methods on biomedical topics. The creation of publicly-available specialized datasets for biomedical FSL may aid method development by facilitating comparative analyses.