Gross failure rates and failure modes for a commercial AI‐based auto‐segmentation algorithm in head and neck cancer patients DOI Creative Commons
Simon W. P. Temple, C. Rowbottom

Journal of Applied Clinical Medical Physics, Journal Year: 2024, Volume and Issue: 25(6)

Published: Jan. 23, 2024

Abstract Purpose Artificial intelligence (AI) based commercial software can be used to automatically delineate organs at risk (OAR), with potential for efficiency savings in the radiotherapy treatment planning pathway, and reduction of inter‐ intra‐observer variability. There has been little research investigating gross failure rates modes such systems. Method 50 head neck (H&N) patient data sets “gold standard” contours were compared AI‐generated produce expected mean standard deviation values Dice Similarity Coefficient (DSC), four common H&N OARs (brainstem, mandible, left right parotid). An AI‐based system was applied 500 patients. manual contours, outlined by an expert human, a set three deviations below DSC. Failures inspected assess reason failures relating suboptimal contouring censored. True classified into 4 sub‐types (setup position, anatomy, image artefacts unknown). Results 24 true software, rate 1.2%. Fifteen due dental artefacts, position two unknown. OAR 0.4% (brainstem), 2.2% (mandible), 1.4% (left parotid) 0.8% (right Conclusion predominantly associated non‐standard element within CT scan. It is likely that these elements failure, suggests datasets train AI model did not contain sufficient heterogeneity data. Regardless reasons region investigated low (∼1%).

Language: Английский

Artificial intelligence and machine learning for medical imaging: A technology review DOI Open Access
Ana María Barragán Montero, Umair Javaid, Gilmer Valdés

et al.

Physica Medica, Journal Year: 2021, Volume and Issue: 83, P. 242 - 256

Published: March 1, 2021

Language: Английский

Citations

275
Artificial intelligence and machine learning in cancer imaging DOI Creative Commons
Dow‐Mu Koh, Nickolas Papanikolaou, Ulrich Bick

et al.

Communications Medicine, Journal Year: 2022, Volume and Issue: 2(1)

Published: Oct. 27, 2022

An increasing array of tools is being developed using artificial intelligence (AI) and machine learning (ML) for cancer imaging. The development an optimal tool requires multidisciplinary engagement to ensure that the appropriate use case met, as well undertake robust testing prior its adoption into healthcare systems. This review highlights key developments in field. We discuss challenges opportunities AI ML imaging; considerations algorithms can be widely used disseminated; ecosystem needed promote growth

Language: Английский

Citations

194
Metrics to evaluate the performance of auto-segmentation for radiation treatment planning: A critical review DOI
Michael V. Sherer, Diana Lin,

Sharif Elguindi

et al.

Radiotherapy and Oncology, Journal Year: 2021, Volume and Issue: 160, P. 185 - 191

Published: May 11, 2021

Language: Английский

Citations

157
Machine Learning for Auto-Segmentation in Radiotherapy Planning DOI
K. Harrison,

H. Pullen,

Ceilidh Welsh

et al.

Clinical Oncology, Journal Year: 2022, Volume and Issue: 34(2), P. 74 - 88

Published: Jan. 5, 2022

Language: Английский

Citations

80
Development and validation of fully automated robust deep learning models for multi-organ segmentation from whole-body CT images DOI Creative Commons
Yazdan Salimi, Isaac Shiri, Zahra Mansouri

et al.

Physica Medica, Journal Year: 2025, Volume and Issue: 130, P. 104911 - 104911

Published: Feb. 1, 2025

This study aimed to develop a deep-learning framework generate multi-organ masks from CT images in adult and pediatric patients. A dataset consisting of 4082 ground-truth manual segmentation various databases, including 300 cases, were collected. In strategy#1, the provided by public databases split into training (90%) testing (10% each database named subset #1) cohort. The set was used train multiple nnU-Net networks five-fold cross-validation (CV) for 26 separate organs. next step, trained models strategy #1 missing organs entire dataset. generated data then model CV (strategy#2). Models' performance evaluated terms Dice coefficient (DSC) other well-established image metrics. lowest DSC strategy#1 0.804 ± 0.094 adrenal glands while average > 0.90 achieved 17/26 strategy#2 (0.833 0.177) obtained pancreas, whereas 13/19 For all mutual included #2, our outperformed TotalSegmentator both strategies. addition, on #3. Our with significant variability different producing acceptable results making it well-suited implementation clinical setting.

Language: Английский

Citations

3
Artificial Intelligence in magnetic Resonance guided Radiotherapy: Medical and physical considerations on state of art and future perspectives DOI Open Access
Davide Cusumano, Luca Boldrini, Jennifer Dhont

et al.

Physica Medica, Journal Year: 2021, Volume and Issue: 85, P. 175 - 191

Published: May 1, 2021

Language: Английский

Citations

79
Review of Deep Learning Based Automatic Segmentation for Lung Cancer Radiotherapy DOI Creative Commons
Xi Liu,

Kai-Wen Li,

Ruijie Yang

et al.

Frontiers in Oncology, Journal Year: 2021, Volume and Issue: 11

Published: July 8, 2021

Lung cancer is the leading cause of cancer-related mortality for males and females. Radiation therapy (RT) one primary treatment modalities lung cancer. While delivering prescribed dose to tumor targets, it essential spare tissues near targets—the so-called organs-at-risk (OARs). An optimal RT planning benefits from accurate segmentation gross volume surrounding OARs. Manual a time-consuming tedious task radiation oncologists. Therefore, crucial develop automatic image relieve oncologists contouring work. Currently, atlas-based technique commonly used in clinical routines. However, this depends heavily on similarity between atlas segmented. With significant advances made computer vision, deep learning as part artificial intelligence attracts increasing attention medical segmentation. In article, we reviewed based techniques related compared them with technique. At present, auto-segmentation OARs relatively large such heart etc. outperforms organs small esophagus. The average Dice coefficient (DSC) lung, liver are over 0.9, best DSC spinal cord reaches 0.9. esophagus ranges 0.71 0.87 ragged performance. terms volume, below 0.8. Although indicate superiority many aspects manual segmentation, various issues still need be solved. We discussed potential including low contrast, dataset size, consensus guidelines, network design. Clinical limitations future research directions were well.

Language: Английский

Citations

78
Anatomy-aided deep learning for medical image segmentation: a review DOI Creative Commons
Lu Liu, Jelmer M. Wolterink, Christoph Brüne

et al.

Physics in Medicine and Biology, Journal Year: 2021, Volume and Issue: 66(11), P. 11TR01 - 11TR01

Published: April 27, 2021

Deep learning (DL) has become widely used for medical image segmentation in recent years. However, despite these advances, there are still problems which DL-based fails. Recently, some DL approaches had a breakthrough by using anatomical information is the crucial cue manual segmentation. In this paper, we provide review of anatomy-aided covers systematically summarized categories and corresponding representation methods. We address known potentially solvable challenges present categorized methodology overview on with from over 70 papers. Finally, discuss strengths limitations current suggest potential future work.

Language: Английский

Citations

64
The impact of training sample size on deep learning-based organ auto-segmentation for head-and-neck patients DOI Creative Commons

Yingtao Fang,

Jiazhou Wang, Xiaomin Ou

et al.

Physics in Medicine and Biology, Journal Year: 2021, Volume and Issue: 66(18), P. 185012 - 185012

Published: Aug. 27, 2021

To investigate the impact of training sample size on performance deep learning-based organ auto-segmentation for head-and-neck cancer patients, a total 1160 patients with who received radiotherapy were enrolled in this study. Patient planning CT images and regions interest (ROIs) delineation, including brainstem, spinal cord, eyes, lenses, optic nerves, temporal lobes, parotids, larynx body, collected. An evaluation dataset 200 randomly selected combined Dice similarity index to evaluate model performances. Eleven datasets different sizes from remaining 960 form models. All models used same data augmentation methods, network structures hyperparameters. A estimation based inverse power law function was established. Different change patterns found organs. Six organs had best 800 samples others achieved their 600 or 400 samples. The benefit increasing gradually decreased. Compared performance, nerves lenses reached 95% effect at 200, other 40. For fitting function, fitted root mean square errors all ROIs less than 0.03 (left eye: 0.024, others: <0.01), theRsquare except body greater 0.5. has significant auto-segmentation. relationship between depends inherent characteristics organ. In some cases, relatively small can achieve satisfactory performance.

Language: Английский

Citations

57
A clinical evaluation of the performance of five commercial artificial intelligence contouring systems for radiotherapy DOI Creative Commons
Paul Doolan,

Stefanie Charalambous,

Yiannis Roussakis

et al.

Frontiers in Oncology, Journal Year: 2023, Volume and Issue: 13

Published: Aug. 4, 2023

Auto-segmentation with artificial intelligence (AI) offers an opportunity to reduce inter- and intra-observer variability in contouring, improve the quality of contours, as well time taken conduct this manual task. In work we benchmark AI auto-segmentation contours produced by five commercial vendors against a common dataset.The organ at risk (OAR) generated solutions (Mirada (Mir), MVision (MV), Radformation (Rad), RayStation (Ray) TheraPanacea (Ther)) were compared manually-drawn expert from 20 breast, head neck, lung prostate patients. Comparisons made using geometric similarity metrics including volumetric surface Dice coefficient (vDSC sDSC), Hausdorff distance (HD) Added Path Length (APL). To assess saved, manually draw correct recorded.There are differences number CT offered each solution study (Mir 99; MV 143; Rad 83; Ray 67; Ther 86), all offering some lymph node levels OARs. Averaged across structures, median vDSCs good for systems favorably existing literature: Mir 0.82; 0.88; 0.86; 0.87; 0.88. All offer substantial savings, ranging between: breast 14-20 mins; neck 74-93 20-26 35-42 mins. The averaged was similar systems: 39.8 43.6 36.6 min; 43.2 45.2 mins.All evaluated high significantly reduced could be used render radiotherapy workflow more efficient standardized.

Language: Английский

Citations

42