Dataset Information

Non-Euclidean classification of medically imaged objects via s-reps.

ABSTRACT: Classifying medically imaged objects, e.g., into diseased and normal classes, has been one of the important goals in medical imaging. We propose a novel classification scheme that uses a skeletal representation to provide rich non-Euclidean geometric object properties. Our statistical method combines distance weighted discrimination (DWD) with a carefully chosen Euclideanization which takes full advantage of the geometry of the manifold on which these non-Euclidean geometric object properties (GOPs) live. Our method is evaluated via the task of classifying 3D hippocampi between schizophrenics and healthy controls. We address three central questions. 1) Does adding shape features increase discriminative power over the more standard classification based only on global volume? 2) If so, does our skeletal representation provide greater discriminative power than a conventional boundary point distribution model (PDM)? 3) Especially, is Euclideanization of non-Euclidean shape properties important in achieving high discriminative power? Measuring the capability of a method in terms of area under the receiver operator characteristic (ROC) curve, we show that our proposed method achieves strongly better classification than both the classification method based on global volume alone and the s-rep-based classification method without proper Euclideanization of non-Euclidean GOPs. We show classification using Euclideanized s-reps is also superior to classification using PDMs, whether the PDMs are first Euclideanized or not. We also show improved performance with Euclideanized boundary PDMs over non-linear boundary PDMs. This demonstrates the benefit that proper Euclideanization of non-Euclidean GOPs brings not only to s-rep-based classification but also to PDM-based classification.

SUBMITTER: Hong J

PROVIDER: S-EPMC4821729 | biostudies-literature | 2016 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Non-Euclidean classification of medically imaged objects via s-reps.

Hong Junpyo J Vicory Jared J Schulz Jörn J Styner Martin M Marron J S JS Pizer Stephen M SM

Medical image analysis 20160219

Classifying medically imaged objects, e.g., into diseased and normal classes, has been one of the important goals in medical imaging. We propose a novel classification scheme that uses a skeletal representation to provide rich non-Euclidean geometric object properties. Our statistical method combines distance weighted discrimination (DWD) with a carefully chosen Euclideanization which takes full advantage of the geometry of the manifold on which these non-Euclidean geometric object properties (G ...[more]

PMID: 26963609

Similar Datasets

Project description:PurposeTo develop a clinically meaningful proximal hamstring tear classification system and to present outcome data for defined subtypes.MethodsRetrospective review was undertaken of patients diagnosed with proximal hamstring tears at a single institution from 2012 to 2019. Images were reviewed by an orthopedic surgeon and musculoskeletal radiologist. Tears were classified as Type 1: partial with subtypes (1A, 1 cm or mild complete tear; 1B, 1-2 cm or full tear with <2 cm retraction), Type 2: complete single-tendon tears with subtypes (2c conjoint tendon only; 2s semimembranosus tendon only); or Type 3: complete tears with >2 cm retraction. Demographics, patient-reported outcome measures including Hip Outcome Score, Activities of Daily Living Subscore (HOS-ADL) and patient satisfaction were evaluated. A poor outcome was defined as HOS-ADL < 80%, and the patient acceptable symptom state (PASS) was defined as HOS-ADL 89.7%.ResultsAt a mean follow-up of 38.6 (range: 12-94) months for 114 patients, distributions were as follows: 18.4% Type 1A, 19.2% Type 1B, 7.8% Type 2c, 3.5% Type 2s, and 50.9% Type 3. Intra-observer and inter-observer reliability had a mean Kappa of 0.985 (95% CI: .956, 1.01) and .905 (95% CI: .895 .915). 66 patients underwent surgery, with 68.97% of them being Type 3. The mean HOS-ADL and PASS rate were higher for operatively treated patients (95%, 93.4%) than for nonoperatively treated patients (81.86%, 44.7%). There were significantly more patients satisfied in the surgery group in both Type 1 and Type 3 tears (P = .046 and P = .049). Body mass index was a significant predictor of a poor outcome in Type 3 tears (P = .039). History of corticosteroid or PRP injection, smoking, and diabetes were not significant predictors of a poor outcome.ConclusionWe present an MRI-based classification system for proximal hamstring injuries with both excellent intra-observer and inter-observer reliability. Outcome measures were improved in patients who underwent surgery.Level of evidenceIV, cohort study: diagnostic case series.

Dataset Information

Non-Euclidean classification of medically imaged objects via s-reps.

Publications

Non-Euclidean classification of medically imaged objects via s-reps.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets