Cascaded deep learning for automated lung cancer tumour burden quantification on [ 18 F]FDG PET/CT. | ScholarDeck

Paper

Cascaded deep learning for automated lung cancer tumour burden quantification on [ 18 F]FDG PET/CT.

New research · Nuclear Medicine

Quantitative imaging in medicine and surgery · 3w

AI / informaticsQuantitative imaging in medicine and surgery· Aug 2026

Cascaded deep learning for automated lung cancer tumour burden quantification on [ 18 F]FDG PET/CT.

Shaonan Zhong, Jie Lv, Zhaohong Pan … Xinlu Wang

Read paper

Oncologic ImagingAI / informaticsDiagnosisQ2IF 2.5

Cascaded AI model matches manual tumor segmentation accuracy on lung cancer PET/CT.

Cascaded deep learning for automated lung cancer tumour burden quantification on [ 18 F]FDG PET/CT.

Shaonan Zhong … Xinlu Wang

Quantitative imaging in medicine and surgery · Aug 2026 · Q2 in Radiology

Background

Accurate assessment of tumour burden in lung cancer is critical for diagnosis, prognosis, and treatment planning.

Methods

In this retrospective single-centre study, we collected 593 fluorine-18 fluorodeoxyglucose ([ 18 F]FDG) PET/CT scans from lung cancer patients with two scanners.

n = 496

Results

0.82

cascaded AI's tumor outlines closely matched expert-drawn ones

n = 496

More results

Tumour burden estimation showed strong correlations with manual measurements (r=0.984 for metabolic tumour volume, r=0.998 for total lesion glycolysis; both P<0.05) and moderate agreement.

“

Conclusion

The proposed cascaded segmentation-classification architecture significantly reduces false positives and yields reliable tumour burden quantification on PET/CT, enhancing accuracy and clinical utility.

Read paper