AItrix, 'Improved Prediction of Patient Condition Deterioration' Selected as Outstanding Paper at MLHC

Medical AI company AITRICS announced on the 18th that its research paper on improving patient deterioration prediction using multi-modal data has been selected as an outstanding paper at the world's largest medical AI conference, the 2023 Machine Learning for Healthcare (MLHC).

Image from AItrics' paper accepted at MLHC 2023
Photo by AItrics

MLHC is the largest conference in the field of AI technology utilizing medical big data. Since 2011, the annual conference has facilitated active exchanges and discussions among AI, machine learning, clinical, and medical experts. This year’s conference will be held on July 11-12 at Columbia University in New York, USA.

The paper by AITRICS selected for MLHC 2023 focuses on addressing issues related to the use of multi-modality in electronic health records (EHR). In existing clinical settings using EHR, various modalities provide rich information, but as the number of modalities increases, computational load rises and data input cycles become irregular.

In this study, AITRICS developed the ▲Unified Multi-Modal Embedding (UMSE) module and ▲Flexible Multi-Modal Learning method, which solved the irregular data input problem using only the original data. Even in environments with some missing data, the model effectively processed and utilized available data to improve patient condition prediction performance. Accordingly, the model developed by AITRICS demonstrated superior performance compared to other existing models in predicting mortality, vasopressor use, and intubation events occurring within 12 hours in patients.

This paper was selected for oral presentation among the submissions to MLHC 2023 and will be presented at the conference next month.

Researcher Kwanhyung Lee of AITRICS stated, “This study confirmed that AI can comprehensively utilize biosignals, X-ray images, and clinical note data in a manner similar to actual medical staff to improve the accuracy of patient condition prediction. Notably, this is the first case of deep fusion of three or more EHR multimodalities.” He added, “The oral presentation is an opportunity to showcase AITRICS’ AI technology and research achievements globally. Moving forward, AITRICS will continue its efforts to help medical professionals effectively identify high-risk patients in clinical settings and make prompt decisions.”

This content was produced with the assistance of AI translation services.