A robust pathological lung segmentation platform using fuzzy—connectedness with patient-specific modeling
Radiological Society of North America (RSNA) 2013 Scientific Assembly and Annual Meeting
Abstract
An abstract is not available in the lab archive. View the paper at its source.
Awais Mansoor and colleagues presented this work at the RSNA 2013 Scientific Assembly and Annual Meeting. The poster describes a CT-based method combining fuzzy-connectedness segmentation, patient-specific shape modeling, and random-forest classification to segment pathological lungs.
Summary
The RSNA abstract describes a proposed automated method for segmenting pathological lungs in computed tomography (CT) scans. It combines fuzzy connectedness, an image-segmentation method, with a patient-specific shape model; regions identified as potential segmentation failures are further refined using a random-forest classifier to label lung tissue as normal or abnormal.
The abstract reports evaluation using 200 CT scans with varying pathology levels and 40 healthy-control scans, and reports an average Dice Similarity Coefficient greater than 90% against expert-radiologist reference standards. These are findings reported in the conference abstract; the source does not establish prospective validation or clinical readiness.
- Presented as a scientific informal poster on December 5, 2013, at the Radiological Society of North America (RSNA) 2013 Scientific Assembly and Annual Meeting.
- RSNA abstract identifier: LL-INS-TH4A.
Cite this work
Awais Mansoor; Ulas Bagci; Brent Foster; Ziyue Xu; Jayaram K. Udupa; Daniel Joseph Mollura (2013). A robust pathological lung segmentation platform using fuzzy—connectedness with patient-specific modeling. Radiological Society of North America (RSNA) 2013 Scientific Assembly and Annual Meeting https://archive.rsna.org/2013/13044303.html
BibTeX