AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

Modeling the Density of Pixel-level Self-supervised Embeddings for Unsupervised Pathology Segmentation in Medical CT

Mikhail Goncharov, Eugenia Soboleva, Daniil Ignatyev, Mariia Donskova, Mikhail Belyaev, Ivan Oseledets, Marina Munkhoeva, Maxim Panov

ICLR 2026 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Pretext-task visual pre-training

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add masking-invariant condition model (vs no condition) to Gaussian density model (ablation)Screener Gaussian density · KiTSvoxel-level AUROC: 0.61 -> 0.87Improved
add masking-invariant condition model (vs no condition) to Gaussian density model (ablation)Screener Gaussian density · LIDCvoxel-level AUROC: 0.81 -> 0.96Improved
replace sin-cos positional encoding with masking-invariant condition (Gaussian) (ablation)Screener Gaussian density · LiTSvoxel-level AUROC: 0.77 -> 0.90Improved
replace APE with masking-invariant condition (Gaussian) (ablation)Screener Gaussian density · LiTSvoxel-level AUROC: 0.86 -> 0.90Improved
add conditioning to normalizing flow density model (ablation)Screener flow density · LIDCvoxel-level AUROC: 0.96 -> 0.96No clear effect
replace ImageNet-pretrained descriptor (MSFlow) with dense SSL descriptor (ablation)density-based UVAS · MIDRCvoxel-level AUROC: 0.66 -> 0.89Improved
replace supervised STU-Net descriptor with DenseVICReg descriptor (ablation)normalizing flow UVAS · MIDRCvoxel-level AUROC: 0.44 -> 0.89Improved
use DenseVICReg instead of DenseInfoNCE descriptor (ablation)normalizing flow UVAS · LiTSvoxel-level AUROC: 0.91 -> 0.93Improved
increase descriptor dimension from 32 to 128 (ablation)DenseVICReg descriptor · KiTSvoxel-level AUROC: 0.88 -> 0.87No clear effect
add Screener pretraining before fine-tuning (25 labeled cases)nnUNet · LIDCDice: 0.21 -> 0.31Improved
add Screener pretraining before fine-tuning (25 labeled cases)nnUNet · KiTSDice: 0.41 -> 0.43No clear effect
add Screener pretraining before fine-tuning on full datannUNet · KiTSDice: 0.78 -> 0.76Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy