Optimising a 3D convolutional neural network for head and neck computed tomography segmentation with limited training data

Henderson, E.; Osorio, Eliana Vásquez; Herk, Marcel van; Green, Andrew

doi:10.1016/j.phro.2022.04.003

Cited by 11 publications

(6 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Pace et al [61] successfully conducted research on the segmentation of medical MRI images in a group of 20 patients with congenital heart disease (CHD) using recursive neu-E a r l y b i r d ral networks. Similar results were achieved by Henderson et al [35] in AI-supported oncology CT analysis in 35 patients. AI-based analysis of small imaging data-sets allows to introduce truly preventive medicine and personalized therapy in rare diseases, including HD.…”

Section: Artificial Intelligence Machine Learning and Deep Learningsupporting

confidence: 88%

Survey on the Most Current Image Processing Methods in Huntington's Disease Diagnostics and Progression Assessment

Kawala-Sterniuk,

Mikolajewski,

Bryniarska

et al. 2024

csci

View full text Add to dashboard Cite

Huntington's disease (HD) is a rare, incurable neurodegenerative disorder where fast and non-invasive diagnosis targeting patients' condition plays a crucial role. In modern medicine, various scientific areas are being combined, such as computing, medicine and biomedical engineering. This survey is focused on the most recent image processing methods applied not only for the purpose of diagnosing HD but also for the assessment of its progression severity, in order to contribute to the effort to prolong life of and to improve its quality.

show abstract

Section: Artificial Intelligence Machine Learning and Deep Learningsupporting

confidence: 88%

Survey on the Most Current Image Processing Methods in Huntington's Disease Diagnostics and Progression Assessment

Kawala-Sterniuk,

Mikolajewski,

Bryniarska

et al. 2024

csci

View full text Add to dashboard Cite

show abstract

“…[ 34 , 53 ]. Edward [ 66 ] performed data augmentation using limited data and evaluated the segmentation effect of a custom model (3D CNN) on a small data set, and the algorithm yielded an average surface distance of only 0.81 mm for the brainstem. Zhao et al [ 67 ] used a principal component analysis model to randomly deform the original CT image to produce new data, and data augmentation provided small-sample-high-quality variants of the contours for DL.…”

Section: Discussionmentioning

confidence: 99%

Deep learning algorithm performance in contouring head and neck organs at risk: a systematic review and single-arm meta-analysis

Liu,

Sun,

Zhao

et al. 2023

BioMed Eng OnLine

View full text Add to dashboard Cite

Purpose The contouring of organs at risk (OARs) in head and neck cancer radiation treatment planning is a crucial, yet repetitive and time-consuming process. Recent studies have applied deep learning (DL) algorithms to automatically contour head and neck OARs. This study aims to conduct a systematic review and meta-analysis to summarize and analyze the performance of DL algorithms in contouring head and neck OARs. The objective is to assess the advantages and limitations of DL algorithms in contour planning of head and neck OARs. Methods This study conducted a literature search of Pubmed, Embase and Cochrane Library databases, to include studies related to DL contouring head and neck OARs, and the dice similarity coefficient (DSC) of four categories of OARs from the results of each study are selected as effect sizes for meta-analysis. Furthermore, this study conducted a subgroup analysis of OARs characterized by image modality and image type. Results 149 articles were retrieved, and 22 studies were included in the meta-analysis after excluding duplicate literature, primary screening, and re-screening. The combined effect sizes of DSC for brainstem, spinal cord, mandible, left eye, right eye, left optic nerve, right optic nerve, optic chiasm, left parotid, right parotid, left submandibular, and right submandibular are 0.87, 0.83, 0.92, 0.90, 0.90, 0.71, 0.74, 0.62, 0.85, 0.85, 0.82, and 0.82, respectively. For subgroup analysis, the combined effect sizes for segmentation of the brainstem, mandible, left optic nerve, and left parotid gland using CT and MRI images are 0.86/0.92, 0.92/0.90, 0.71/0.73, and 0.84/0.87, respectively. Pooled effect sizes using 2D and 3D images of the brainstem, mandible, left optic nerve, and left parotid gland for contouring are 0.88/0.87, 0.92/0.92, 0.75/0.71 and 0.87/0.85. Conclusions The use of automated contouring technology based on DL algorithms is an essential tool for contouring head and neck OARs, achieving high accuracy, reducing the workload of clinical radiation oncologists, and providing individualized, standardized, and refined treatment plans for implementing "precision radiotherapy". Improving DL performance requires the construction of high-quality data sets and enhancing algorithm optimization and innovation.

show abstract

“…Henderson et al [3] used a publicly available CT dataset of 35 patients [5] , and assessed the influence of using one or three input channels with different window level settings (soft-tissue, bone, brain), three different loss functions (multi-class weighted soft-dice (wSD), cross-entropy (XE) + wSD, and Exponential Logarithmic Loss (ExpLogLoss), and the use of transpose vs resize convolutions in the up-sampling part of their convolutional neural network (CNN). For external validation, they also took their optimal model configuration and trained with the 2015 OAR challenge dataset, using the 25 patients of the training set for training, the 5 onsite testing patients for validation, and the 10 offsite patients for testing.…”

Section: Oar Segmentation In Hncmentioning

confidence: 99%

“…Themes that often are addressed in DLS research evolve around the impact of dataset size, dealing with the class imbalance problem, and especially for GTV segmentation, uni- vs multi-modal imaging. Recently, two papers in the area of DLS in HNC were published in Physics and Imaging in Radiation Oncology , by Henderson et al [3] and Outeiral et al [4] . The Henderson et al paper focused on OAR segmentation (brainstem, mandible, parotid glands, spinal cord) on computed tomography (CT) scans, while the latter addressed primary tumor segmentation in the oropharynx on magnetic resonance imaging (MRI)-only data.…”

mentioning

confidence: 99%

Challenges and chances for deep-learning based target and organ at risk segmentation in radiotherapy of head and neck cancer

Nijkamp

2022

Physics and Imaging in Radiation Oncology

View full text Add to dashboard Cite

Optimising a 3D convolutional neural network for head and neck computed tomography segmentation with limited training data

Cited by 11 publications

References 29 publications

Survey on the Most Current Image Processing Methods in Huntington's Disease Diagnostics and Progression Assessment

Survey on the Most Current Image Processing Methods in Huntington's Disease Diagnostics and Progression Assessment

Deep learning algorithm performance in contouring head and neck organs at risk: a systematic review and single-arm meta-analysis

Challenges and chances for deep-learning based target and organ at risk segmentation in radiotherapy of head and neck cancer

Contact Info

Product

Resources

About