Unbalanced optimal transport for robust longitudinal lesion evolution with registration-aware and appearance-guided priors
View original at arxiv.org{ "id": "2602.09933v1", "url": "http://arxiv.org/abs/2602.09933v1", "title": "Unbalanced optimal transport for robust longitudinal lesion evolution with registration-aware and appearance-guided priors", "summary": "Evaluating lesion evolution in longitudinal CT scans of can cer patients is essential for assessing treat…
What we drew from this source
The claims Via News extracted from this document. We point to the source; we don't replace it.
Accurate detection of merging and splitting lesions is crucial for reliable response evaluation, as overlooking these events can lead to misclassification under RECIST and potentially incorrect assessment of disease progression
80% confidenceStandard bipartite matchers which rely on geometric proximity struggle when lesions appear, disappear, merge, or split
80% confidenceThis is the first approach to cast longitudinal lesion correspondence as a UOT problem, providing a principled alternative to distance-based bipartite matchers
80% confidenceThe proposed method produces interpretable lesion evolution graph with persistent, new, disappearing, merging, and splitting events without requiring heuristics or training data
80% confidenceUOT achieves consistently higher edge-detection precision and recall, improved lesion-state recall, and superior lesion-graph component F1 scores versus distance-only baselines
80% confidence
Cited in these Via News reports
- Computer Vision AI Deploys in High-Risk Medicine and Robotics Across Five Continents →
- Computer Vision AI Enters Commercial Phase as Ethics Concerns Challenge Big Tech's Dominance →
- Computer Vision AI Moves to Edge Processing as Healthcare and Telecom Deploy Privacy-First Systems Globally →
- Computer Vision AI Ships in Consumer Hardware Across US, Europe, Asia Markets as Edge Computing Tackles Clinical Imaging Delays →
- Computer Vision Splits Into Specialized Systems as Medical, Industrial Apps Reject Universal Models →
- Computer Vision Splits Into Two Markets: Mobileye's $1.9B Auto Play Stalls While Factory Tools Surge →
- Enterprise AI Moves to Edge Devices as Privacy Rules and Latency Needs Force Global Shift from Cloud →
- Medical AI Diagnostics Deploy Across US, Europe as Automotive Vision Merges with Robotics →
- Medical AI Radiology Systems Enter Clinical Use in U.S. and Europe as Automotive Vision Expands to Robotics Control →
- Medical Imaging AI Systems Deploy Across Global Healthcare Networks After Validation Trials →
- Meta's 200-Language Model Triggered African AI Startup Closures, Gebru Says →
- Meta's Universal AI Model Collapsed African Language Startup Funding, Says AI Researcher →