Lung Nodule Detection Via Bayesian Voxel Labeling

Mendonça, Paulo R. S.; Bhotika, Rahul; Zhao, Fei; Miller, James V.

doi:10.1007/978-3-540-73273-0_12

“…We validate its effectiveness by evaluating the impacts on large-scale colon and lung CAD system performances (879 and 770 volumes respectively). The results are very encouraging and significantly outperform the recent state-of-the-arts [1,2,5,[11][12][13][14][15].…”

Section: Introductionmentioning

confidence: 72%

“…On the other end, [4] utilizes analytical shape and appearance priors and Markov-Gibbs random field; [1] designs elaborate region-growing criteria separating nodule growth from normal tissues; and [3] empowers morphological approaches and convexity models. CAD detection bias or dependency on segmentation (e.g., underor over-segmentation often occurred, segmentation failures) may not be desirable as discussed in [5]. A Bayesian voxel labeling approach for lung nodule detection was thus proposed [5], avoiding explicit segmentation.…”

Section: Introductionmentioning

confidence: 99%

“…CAD detection bias or dependency on segmentation (e.g., underor over-segmentation often occurred, segmentation failures) may not be desirable as discussed in [5]. A Bayesian voxel labeling approach for lung nodule detection was thus proposed [5], avoiding explicit segmentation. Nevertheless, their four types of probabilistic formulations of nodule, vessel, vessel junction and outlier are based on medical literature description or general knowledge, but not verified from a large amount of data.…”

Section: Introductionmentioning

confidence: 99%

“…Their detailed model parameters are heuristically decided from common medical prior knowledge, for example, 15mm for the average maximum radius of a pulmonary vessel. The nodule model in [5] is invalid for partial-solid and ground-glass nodules which nevertheless have great importance for CAD to perform well, as it is more ambiguous for radiologists. The solid parenchymal nodule model may also have trouble with other contextual types (e.g., juxta-pleural, pleural tail, and vascular [6]).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Computer Aided Diagnosis Using Multilevel Image Features on Large-Scale Evaluation

Lü

¹

,

Devarakota²,

Vikal³

et al. 2014

Medical Computer Vision. Large Data in Medical Imaging

View full text Add to dashboard Cite

Abstract. Computer aided diagnosis (CAD) of cancerous anatomical structures via 3D medical images has emerged as an intensively studied research area. In this paper, we present a principled three-tiered image feature learning approach to capture task specific and data-driven class discriminative statistics from an annotated image database. It integrates voxel-, instance-, and database-level feature learning, aggregation and parsing. The initial segmentation is proceeded as robust voxel labeling and thresholding. After instance-level spatial aggregation, extracted features can also be flexibly tuned for classifying lesions, or discriminating different subcategories of lesions. We demonstrate the effectiveness in the lung nodule detection task which handles all types of solid, partial-solid, and ground-glass nodules using the same set of learned features. Our hierarchical feature learning framework, which was extensively trained and validated on large-scale multiple site datasets of 879 CT volumes (510 training and 369 validation), achieves superior performance than other state-of-the-art CAD systems. The proposed method is also shown to be applicable for colonic polyp detection, including all polyp morphological subcategories, via 770 tagged-prep CT scans from multiple medical sites (358 training and 412 validation).

show abstract

“…However, we could account for colonic haustra variations in size and shape by marginalizing over the parameter R. The prior for this marginalization will depend upon either training data of prior clinical knowledge of insufflated haustra radii. In [15] we have shown how to compute the probability distribution of curvatures for a class of ellipsoidal surfaces, suggesting a mechanism to achieve such generalization.…”

Section: Discussionmentioning

confidence: 99%

A Probabilistic Model for Haustral Curvatures with Applications to Colon CAD

Melonakos

¹

,

Mendonça

²

,

Bhotka

³

et al.

Medical Image Computing and Computer-Assisted Intervention – MICCAI 2007

View full text Add to dashboard Cite

Abstract. Among the many features used for classification in computer-aided detection (CAD) systems targeting colonic polyps, those based on differences between the shapes of polyps and folds are most common. We introduce here an explicit parametric model for the haustra or colon wall. The proposed model captures the overall shape of the haustra and we use it to derive the probability distribution of features relevant to polyp detection. The usefulness of the model is demonstrated through its application to a colon CAD algorithm.

show abstract

Computer Aided Diagnosis Using Multilevel Image Features on Large-Scale Evaluation

Lü

¹

,

Devarakota²,

Vikal³

et al. 2014

Medical Computer Vision. Large Data in Medical Imaging

View full text Add to dashboard Cite

Abstract. Computer aided diagnosis (CAD) of cancerous anatomical structures via 3D medical images has emerged as an intensively studied research area. In this paper, we present a principled three-tiered image feature learning approach to capture task specific and data-driven class discriminative statistics from an annotated image database. It integrates voxel-, instance-, and database-level feature learning, aggregation and parsing. The initial segmentation is proceeded as robust voxel labeling and thresholding. After instance-level spatial aggregation, extracted features can also be flexibly tuned for classifying lesions, or discriminating different subcategories of lesions. We demonstrate the effectiveness in the lung nodule detection task which handles all types of solid, partial-solid, and ground-glass nodules using the same set of learned features. Our hierarchical feature learning framework, which was extensively trained and validated on large-scale multiple site datasets of 879 CT volumes (510 training and 369 validation), achieves superior performance than other state-of-the-art CAD systems. The proposed method is also shown to be applicable for colonic polyp detection, including all polyp morphological subcategories, via 770 tagged-prep CT scans from multiple medical sites (358 training and 412 validation).

show abstract

Lung Nodule Detection Via Bayesian Voxel Labeling

Cited by 15 publications

References 27 publications

Computer Aided Diagnosis Using Multilevel Image Features on Large-Scale Evaluation

Computer Aided Diagnosis Using Multilevel Image Features on Large-Scale Evaluation

A Probabilistic Model for Haustral Curvatures with Applications to Colon CAD

Computer Aided Diagnosis Using Multilevel Image Features on Large-Scale Evaluation

Contact Info

Product

Resources

About