{"paper":{"title":"Uncertainty-Aware Image Classification In Biomedical Imaging Using Spectral-normalized Neural Gaussian Processes","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Spectral-normalized neural Gaussian processes deliver accurate biomedical image classification with improved uncertainty estimates for out-of-distribution inputs.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Jeffrey J. Nirschl, Uma Meleti","submitted_at":"2026-02-02T17:35:10Z","abstract_excerpt":"Accurate histopathologic interpretation is key for clinical decision-making; however, current deep learning models for digital pathology are often overconfident and poorly calibrated in out-of-distribution (OOD) settings, which limit trust and clinical adoption. Safety-critical medical imaging workflows benefit from intrinsic uncertainty-aware properties that can accurately reject OOD input. We implement the Spectral-normalized Neural Gaussian Process (SNGP), a set of lightweight modifications that apply spectral normalization and replace the final dense layer with a Gaussian process layer to "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"SNGP has comparable in-distribution performance while significantly improving uncertainty estimation and OOD detection.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The chosen OOD test sets accurately represent the distribution shifts that occur in real clinical pathology workflows.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SNGP models match deterministic neural network accuracy on biomedical images while providing superior uncertainty calibration and OOD rejection across six datasets.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Spectral-normalized neural Gaussian processes deliver accurate biomedical image classification with improved uncertainty estimates for out-of-distribution inputs.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"8e42bc303c86c5415a237fb1be98b239d55ff6802f159f74bdb5c95df56e9221"},"source":{"id":"2602.02370","kind":"arxiv","version":2},"verdict":{"id":"a1d03a16-5d74-422f-860f-a0e9f8d4735c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T08:01:01.660921Z","strongest_claim":"SNGP has comparable in-distribution performance while significantly improving uncertainty estimation and OOD detection.","one_line_summary":"SNGP models match deterministic neural network accuracy on biomedical images while providing superior uncertainty calibration and OOD rejection across six datasets.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The chosen OOD test sets accurately represent the distribution shifts that occur in real clinical pathology workflows.","pith_extraction_headline":"Spectral-normalized neural Gaussian processes deliver accurate biomedical image classification with improved uncertainty estimates for out-of-distribution inputs."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2602.02370/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":17,"sample":[{"doi":"","year":2026,"title":"Uncertainty-Aware Image Classification In Biomedical Imaging Using Spectral-normalized Neural Gaussian Processes","work_id":"f15a4692-fb67-4c75-9857-ce4512cdcc4a","ref_index":1,"cited_arxiv_id":"2602.02370","is_internal_anchor":true},{"doi":"","year":null,"title":"Bayesian neural networks provide a formal approach to uncertainty estimation but are compu- tationally impractical for large architectures","work_id":"406c5ffe-4e42-433d-b304-0e1e7d31a4b9","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Datasets All datasets are publicly available from the original authors or MicroBench [10]","work_id":"1cb54343-de3d-4524-bf7d-d88d3ee7c197","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"RESULTS Tab.1 summarizes the OOD detection performance of all methods trained on the Acevedo dataset. SNGP achieved near-perfect OOD-AUROC across all external OOD datasets (0.97–1.00), while maintaini","work_id":"0d2ebe3e-a63b-4c39-a15e-d0fd46c53220","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Across multiple datasets, it maintains strong calibration and in- distribution accuracy while substantially improving OOD detection over deterministic and Monte Carlo methods","work_id":"881ffbbb-cf95-4656-a7e1-6138ab7c6807","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":17,"snapshot_sha256":"b027b6f9e2fa003d167c8567eb0977f0f110c175e831700b0c88424d19362de4","internal_anchors":1},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}