Location registration and recognition (LRR) for serial analysis of nodules in lung CT scans.

摘要: In the clinical workflow for lung cancer management, comparison of nodules between CT scans from subsequent visits by a patient is necessary timely classification pulmonary into benign and malignant analyzing nodule growth response to therapy. The algorithm described in this paper takes (a) two temporally-separated scans, I(1) I(2), (b) series locations I(1), each location it produces an affine transformation that maps their immediate neighborhoods I(2). It does without deformable registration initialization global registration. Requiring be specified only one volume provides clinician more flexibility investigating condition lung. uses combination feature extraction, indexing, refinement, decision processes. Together, these processes essentially "recognize" neighborhoods. We show on our technique works at near interactive speed median alignment error 134 1.70mm compared 2.14mm Diffeomorphic Demons algorithm, 3.57mm with local refinement. demonstrate 250 nodules, robust changes caused progression differences breathing states, scanning procedures, positioning. Our may used both diagnosis treatment monitoring cancer. Because generic design might also other applications require fast accurate mapping regions.