Advanced breath analysis through hierarchical deep convolutional neural network for multi-cancer screening

1

Cancer accounts for over 16% of global deaths, with lung and gastric cancers being among the most lethal malignancies.
2

Breath analysis offers a noninvasive method for cancer detection by capturing volatile organic compounds as disease-specific biomarkers.
3

Current studies primarily focus on single-cancer detection, highlighting the need for systems capable of identifying multiple cancers simultaneously.
4

The proposed hierarchical deep convolutional neural network (HD-CNN) enhances multi-cancer classification accuracy from breath samples.
5

The upgraded breath analysis system features a multimodal sensor array and improved thermal control for reliable VOC detection.

Npj Digital Medicine

by Byeongju Lee, Junyeong Lee, Hyowoong Noh, Hyung-Keun Bahn, Jae-Hyun Jeon, Inkyu Park, Sanghoon Jheon, Dae-Sik Lee
January 8, 2026

1