Biotechnology

Deep learning chest radiograph model predicts mortality for


Leesburg, VA, June 30, 2023—according to the accepted manuscript published by ARRS itself American Journal of Roentgenology (AIR), a deep learning-based model using baseline chest radiographs predicts 30-day mortality in patients with community-acquired pneumonia (CAP), improving the performance of established risk predictor tools (ie, CURB-65 score).

Leesburg, VA, June 30, 2023—according to the accepted manuscript published by ARRS itself American Journal of Roentgenology (AIR), a deep learning-based model using baseline chest radiographs predicts 30-day mortality in patients with community-acquired pneumonia (CAP), improving the performance of established risk predictor tools (ie, CURB-65 score).

“Deep learning models (DL) can guide clinical decision-making in the management of CAP patients by identifying high-risk patients requiring hospitalization and intensive care,” the first authors concluded. Eui Jin HwangMD, PhD, from the radiology department at Seoul National University College of Medicine in Korea.

In the AIR accepted manuscripts, the DL model was developed in 7,105 patients through a single institution from March 2013 to December 2019 (3:1:1 allocation for training, validation, and internal test set) to predict the risk of all-cause death within 30 days of CAP diagnosis using X-ray patient’s initial chest. Hwang et al. then evaluated their DL model in patients diagnosed with CAP during emergency department visits at the same institution as the development cohort from January 2020 to December 2020 (interim trial cohort (n = 947)), and from two additional different institutions (external trial cohort). A (n = 467), January 2020 to December 2020; external test group B (n = 381), March 2019 to October 2021). AUC was compared between the DL models and risk scores based on confusion, blood urea nitrogen levels, respiratory rate, blood pressure, and age ≥ 65 years.

Ultimately, the DL model using baseline chest radiography predicted 30-day all-cause mortality in patients with CAP with an AUC ranging from 0.77 to 0.80 in test cohorts from different institutions. In addition, the models showed higher specificity (range, 61–69%) than CURB-65 scores (44–58%) at the same sensitivity (all p < .001).

Electronic supplement for this AIR accepted manuscripts are available Here.


first radiology society in North America, the American Ray Roentgen Society (ARRS) remains dedicated to the advancement of medicine through the medical imaging profession and its related disciplines. An international forum for the advancement of radiology since the invention of the x-ray, ARRS maintains its mission to improve health through a community committed to advancing knowledge and skills with the world’s longest-running continuously published radiology journal—American Journal of Roentgenology—ARRS Annual Meeting, In practice magazines, topical symposiums, tons of multimedia educational materials, as well as awarding scholarships through Roentgen Fund®.

MEDIA CONTACT:

Logan K. Young, PIO

44211 Slate Field

Leesburg, VA 20176

(email protected)




Source link

Related Articles

Back to top button