Automatic tissue segmentation of head and neck MR images for hyperthermia treatment planning

Fortunati, Valerio; Verhaart, René F.; Niessen, Wiro J.; Veenland, Jifke F.; Paulides, Margarethus M.; Walsum, Theo van

doi:10.1088/0031-9155/60/16/6547

Cited by 24 publications

(36 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Alternative HaN segmentation approaches applied fast marching, mass springs for lymph nodes modeling parametrical shapes for eye globes and optical nerve modeling, active contours, deformable meshes, principal component‐based shapes, graph cut, and superpixels . In general, shape‐based approaches demonstrated exceptional performance on segmentation of heart and spine, and annotation of X‐ray head images .…”

Section: Introductionmentioning

confidence: 99%

“…Researchers also studied the performance of machine learning algorithms such as k‐nearest neighbors and support vector machines trained on intensity, gradient, texture contrast, texture homogeneity, texture energy, cluster tendency, Gabor and Sobel features . The above mentioned algorithms covered the complete set of OARs in the HaN region including brainstem, cerebellum, spinal cord, mandible, parotid glan‐ds, submandibular glands, pituitary gland, thyroid, eye globes, eye lenses, optic nerves, optic chiasm, larynx, pharyngeal constrictor muscle, pterygoid muscles, tongue muscles, and lymph nodes . Despite the considerable attention, the demonstrated results are still not satisfactory for the clinical usage and automated methods cannot accurately segment OARs in the presence of tumors and severe pathologies.…”

Section: Introductionmentioning

confidence: 99%

“…20,27,28 Landmarkassisted atlasing demonstrated better performance in comparison to standard atlasing on segmentation of lymph nodes, 20 brainstem, 27 and individual tongue muscles. 28 Alternative HaN segmentation approaches applied fast marching, 29 mass springs for lymph nodes modeling 30 parametrical shapes for eye globes and optical nerve modeling, 31 active contours, 32 deformable meshes, 24,33 principal component-based shapes, 11,19 graph cut, 13,17 and superpixels. 15 In general, shape-based approaches demonstrated exceptional performance on segmentation of heart 34 and spine, [35][36][37] and annotation of X-ray head images.…”

Section: Introductionmentioning

confidence: 99%

“…38,39 Researchers also studied the performance of machine learning algorithms such as k-nearest neighbors and support vector machines trained on intensity, gradient, texture contrast, texture homogeneity, texture energy, cluster tendency, Gabor and Sobel features. 21,24 The above mentioned algorithms covered the complete set of OARs in the HaN region including brainstem, 4,7-12,14,16-18,24-27,40,41 cerebellum, 9,13,14,17,25,41 spinal cord, 4,13,16,18,40,41 mandible, 4,10,18,24,[42][43][44][45] parotid glands, 4,7,10,11,15,16,18,21,24,33,40,44,46 submandibular glands, 4,24,46 pituitary gland, 9 thyroid, 13,41,47 eye globes, 9,13,14,17,31,40,41 eye lenses, 13,14,16,17,…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Segmentation of organs‐at‐risks in head and neck CT images using convolutional neural networks

2017

View full text Add to dashboard Cite

Purpose Accurate segmentation of organs-at-risks (OARs) is the key step for efficient planning of radiation therapy for head and neck (HaN) cancer treatment. In the work, we proposed the first deep learning-based algorithm, for segmentation of OARs in HaN CT images, and compared its performance against state-of-the-art automated segmentation algorithms, commercial software and inter-observer variability. Methods Convolutional neural networks (CNNs) – a concept from the field of deep learning – were used to study consistent intensity patterns of OARs from training CT images and to segment the OAR in a previously unseen test CT image. For CNN training, we extracted a representative number of positive intensity patches around voxels that belong to the OAR of interest in training CT images, and negative intensity patches around voxels that belong to the surrounding structures. These patches then passed through a sequence of CNN layers that captured local image features such as corners, end-points and edges, and combined them into more complex high-order features that can efficiently describe the OAR. The trained network was applied to classify voxels in a region of interest in the test image where the corresponding OAR is expected to be located. We then smoothed the obtained classification results by using Markov random fields algorithm. We finally extracted the largest connected component of the smoothed voxels classified as the OAR by CNN, performed dilate-erode operations to remov cavities of the component, which resulted in segmentation of the OAR in the test image. Results The performance of CNNs was validated on segmentation of spinal cord, mandible, parotid glands, submandibular glands, larynx, pharynx, eye globes, optic nerves and optic chiasm using 50 CT images. The obtained segmentation results varied from 37.4% Dice coefficient (DSC) for chiasm to 89.5% DSC for mandible. We also analyzed the performance of state-of-the-art algorithms and commercial software reported in the literature, and observed that CNNs demonstrate similar or superior performance on segmentation of spinal cord, mandible, parotid glands, larynx, pharynx, eye globes and optic nerves, but inferior performance on segmentation of submandibular glands and optic chiasm. Conclusion We concluded that convolution neural networks can accurately segment most of OARs using a representative database of 50 HaN CT images. At the same time, inclusion of additional information, e.g. MR images, may be beneficial for some OARs with poorly-visible boundaries.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Segmentation of organs‐at‐risks in head and neck CT images using convolutional neural networks

2017

View full text Add to dashboard Cite

show abstract

“…Patient specific 3D models were automatically segmented from on average 166 ± 44 computer tomography (CT) slices using an atlas-based method [17,18]. Adjustment of air at nose and ear cavities, metal implants and assignment of tumour from radiotherapy planning delineation was done manually in iSeg (v. 3.8, ZMT, Z€ urich, Switzerland).…”

Section: Patient Modelmentioning

confidence: 99%

Hyperthermia treatment planning guided applicator selection for sub-superficial head and neck tumors heating

Drizdal

Paulides

Holthe

et al. 2017

International Journal of Hyperthermia

View full text Add to dashboard Cite

Purpose: In this study, we investigated the differences in hyperthermia treatment (HT) quality between treatments applied with different hyperthermia systems for sub-superficial tumours in the head and neck (H&N) region. Materials and methods: In 24 patients, with a clinical target volume (CTV) extending up to 6 cm from the surface, we retrospectively analysed the predicted HT quality achievable by two planar applicator arrays or one phased-array hyperthermia system. Hereto, we calculated and compared the specific absorption rate (SAR) and temperature distribution coverage of the CTV and gross tumour volume (GTV) for the Lucite cone applicator (LCA: planar), current sheet applicator (CSA: planar) and the HYPERcollar (phased-array). Results: The HYPERcollar provides better SAR coverage than planar applicators if the target region is fully enclosed by its applicator frame. For targets extending outside the HYPERcollar frame, sufficient SAR coverage (25% target coverage, i.e. TC25 ! 75%) can still be achieved using the LCA when the target is fully under the LCA aperture and not deeper than 50 mm from the patient surface. Conclusion: Simulations predict that the HYPERcollar (hence also its successor the HYPERcollar3D) is to be preferred over planar applicators such as LCA and current sheet applicator in sub-superficial tumours in the H&N region when used within specifications. ARTICLE HISTORY

show abstract

AnatomyNet: Deep learning for fast and fully automated whole‐volume segmentation of head and neck anatomy

Zhu

Huang

Zeng³

et al. 2018

Medical Physics

430

333

View full text Add to dashboard Cite

Purpose Radiation therapy (RT) is a common treatment option for head and neck (HaN) cancer. An important step involved in RT planning is the delineation of organs‐at‐risks (OARs) based on HaN computed tomography (CT). However, manually delineating OARs is time‐consuming as each slice of CT images needs to be individually examined and a typical CT consists of hundreds of slices. Automating OARs segmentation has the benefit of both reducing the time and improving the quality of RT planning. Existing anatomy autosegmentation algorithms use primarily atlas‐based methods, which require sophisticated atlas creation and cannot adequately account for anatomy variations among patients. In this work, we propose an end‐to‐end, atlas‐free three‐dimensional (3D) convolutional deep learning framework for fast and fully automated whole‐volume HaN anatomy segmentation. Methods Our deep learning model, called AnatomyNet, segments OARs from head and neck CT images in an end‐to‐end fashion, receiving whole‐volume HaN CT images as input and generating masks of all OARs of interest in one shot. AnatomyNet is built upon the popular 3D U‐net architecture, but extends it in three important ways: (a) a new encoding scheme to allow autosegmentation on whole‐volume CT images instead of local patches or subsets of slices, (b) incorporating 3D squeeze‐and‐excitation residual blocks in encoding layers for better feature representation, and (c) a new loss function combining Dice scores and focal loss to facilitate the training of the neural model. These features are designed to address two main challenges in deep learning‐based HaN segmentation: (a) segmenting small anatomies (i.e., optic chiasm and optic nerves) occupying only a few slices, and (b) training with inconsistent data annotations with missing ground truth for some anatomical structures. Results We collected 261 HaN CT images to train AnatomyNet and used MICCAI Head and Neck Auto Segmentation Challenge 2015 as a benchmark dataset to evaluate the performance of AnatomyNet. The objective is to segment nine anatomies: brain stem, chiasm, mandible, optic nerve left, optic nerve right, parotid gland left, parotid gland right, submandibular gland left, and submandibular gland right. Compared to previous state‐of‐the‐art results from the MICCAI 2015 competition, AnatomyNet increases Dice similarity coefficient by 3.3% on average. AnatomyNet takes about 0.12 s to fully segment a head and neck CT image of dimension 178 × 302 × 225, significantly faster than previous methods. In addition, the model is able to process whole‐volume CT images and delineate all OARs in one pass, requiring little pre‐ or postprocessing. Conclusion Deep learning models offer a feasible solution to the problem of delineating OARs from CT images. We demonstrate that our proposed model can improve segmentation accuracy and simplify the autosegmentation pipeline. With this method, it is possible to delineate OARs of a head and neck CT within a fraction of a second.

show abstract

Automatic tissue segmentation of head and neck MR images for hyperthermia treatment planning

Cited by 24 publications

References 37 publications

Segmentation of organs‐at‐risks in head and neck CT images using convolutional neural networks

Segmentation of organs‐at‐risks in head and neck CT images using convolutional neural networks

Hyperthermia treatment planning guided applicator selection for sub-superficial head and neck tumors heating

AnatomyNet: Deep learning for fast and fully automated whole‐volume segmentation of head and neck anatomy

Contact Info

Product

Resources

About