Closed loop text guided framework for lung cancer lesion segmentation and quantification - Takeaways - MDSpire

Closed loop text guided framework for lung cancer lesion segmentation and quantification

  • By

  • Shiyang Wang

  • Ziyi Wang

  • Wanfu Men

  • Zhenyu Song

  • Dayu Hu

  • Tianyu Liu

  • Boyang Wang

  • Dexing Kong

  • Xuehao Li

  • Kaiming Ren

  • Mingrui Shao

  • February 12, 2026

  • 0 min

Share

  • 1

    BiomedLoop is a text-driven framework that enhances lung cancer lesion segmentation by integrating semantic descriptions with spatial quantification.

  • 2

    The framework utilizes fine-tuned Grounding DINO for localization and SEEM for refinement, supported by an Uncertainty Aware Feature Modulator.

  • 3

    BiomedLoop innovatively converts geometric descriptors from masks into structured pseudo text prompts to improve localization accuracy.

  • 4

    The system generates structured reports that comply with TID 1500 specifications, facilitating alignment with clinical reporting standards.

  • 5

    Experimental results show BiomedLoop outperforms conventional CNNs and Segment Anything Model variants in Dice similarity and Hausdorff distances.

Original Source(s)

Closed loop text guided framework for lung cancer lesion segmentation and quantification

  • by Shiyang Wang, Ziyi Wang, Wanfu Men, Zhenyu Song, Dayu Hu, Tianyu Liu, Boyang Wang, Dexing Kong, Xuehao Li, Kaiming Ren, Mingrui Shao

  • February 12, 2026

Related Content

Severe IL-6–dominant immune-mediated inflammatory response complicated by probable arginine vasopressin deficiency following ivonescimab-based therapy in advanced EGFR-mutant lung squamous cell carcinoma: a case report

Dosimetric and clinical outcomes of stereotactic body radiotherapy for primary lung cancer: isocenter-based vs. volume-based prescription

Gene amplification in the premalignant stages of non-small cell lung cancer development