Artificial intelligence uncertainty quantification in radiotherapy applications − A scoping review

Radiotherapy and Oncology, Journal Year: 2024, Volume and Issue: 201, P. 110542 - 110542

Published: Sept. 17, 2024

Language: Английский

---

Cardiac substructure delineation in radiation therapy – A state‐of‐the‐art review

Journal of Medical Imaging and Radiation Oncology, Journal Year: 2024, Volume and Issue: unknown

Published: May 17, 2024

Summary Delineation of cardiac substructures is crucial for a better understanding radiation‐related cardiotoxicities and to facilitate accurate precise dose calculation developing applying risk models. This review examines recent advancements in substructure delineation the radiation therapy (RT) context, aiming provide comprehensive overview current level knowledge, challenges future directions this evolving field. Imaging used RT planning presents reliably visualising anatomy. Although atlases contouring guidelines aid standardisation reduction variability, significant uncertainties remain defining Coupled with inherent complexity heart, necessitates auto‐contouring consistent large‐scale data analysis improved efficiency prospective applications. Auto‐contouring models, developed primarily breast lung cancer RT, have demonstrated performance comparable manual contouring, marking milestone evolution practices. Nevertheless, several key concerns require further investigation. There an unmet need expanding models encompass broader range sites. A shift focus needed from ensuring accuracy enhancing robustness accessibility Addressing these paramount integration associated into routine clinical practice, thereby improving safety patients.

Language: Английский

---

Effective Adoption of Artificial Intelligence in Healthcare: A Multiple Case Study

Journal of Decision System, Journal Year: 2025, Volume and Issue: 34(1)

Published: Jan. 2, 2025

Language: Английский

---

Performance of Commercial Deep Learning-Based Auto-Segmentation Software for Prostate Cancer Radiation Therapy Planning: A Systematic Review

Information, Journal Year: 2025, Volume and Issue: 16(3), P. 215 - 215

Published: March 11, 2025

As yet, there is no systematic review focusing on benefits and issues of commercial deep learning-based auto-segmentation (DLAS) software for prostate cancer (PCa) radiation therapy (RT) planning despite that NRG Oncology has underscored such necessity. This article’s purpose to systematically DLAS product performances PCa RT their associated evaluation methodology. A literature search was performed with the use electronic databases 7 November 2024. Thirty-two articles were included as per selection criteria. They evaluated 12 products (Carina Medical LLC INTContour (Lexington, KY, USA), Elekta AB ADMIRE (Stockholm, Sweden), Limbus AI Inc. Contour (Regina, SK, Canada), Manteia Technologies Co. AccuContour (Jian Sheng, China), MIM Software ProtégéAI (Cleveland, OH, Mirada Ltd. DLCExpert (Oxford, UK), MVision.ai Contour+ (Helsinki, Finland), Radformation AutoContour (New York, NY, RaySearch Laboratories RayStation Siemens Healthineers AG AI-Rad Companion Organs RT, syngo.via Image Suite DirectORGANS (Erlangen, Germany), Therapanacea Annotate (Paris, France), Varian Systems, Ethos (Palo Alto, CA, USA)). Their results illustrate can delineate organs at risk (abdominopelvic cavity, anal canal, bladder, body, cauda equina, left (L) right (R) femurs, L R pelvis, proximal sacrum) four clinical target volumes (prostate, lymph nodes, bed, seminal vesicle bed) clinically acceptable outcomes, resulting in delineation time reduction, 5.7–81.1%. Although recommended each centre perform its own prior implementation, seems more important due methodological respective single studies, e.g., small dataset used, etc.

Language: Английский

---

Proposal for a Method for Assessing the Quality of an Updated Deep Learning-Based Automatic Segmentation Program

Cureus, Journal Year: 2025, Volume and Issue: unknown

Published: March 27, 2025

This study aimed to verify whether a commercial deep learning-based automatic segmentation (DLS) method can maintain contour geometric accuracy post-update and propose streamlined validation that minimizes the burden on clinical workflows. included 109 participants. Radiation oncologists used computed tomography (CT) imaging identify 28 organs located in head neck, chest, abdomen, pelvic regions. Contours were delineated CT images using AI-Rad Companion Organs RT (AIRC; Siemens Healthineers, Erlangen, Germany) versions VA30, VA50, VA50. The Dice similarity coefficient, maximum Hausdorff distance, mean distance agreement calculated contours with significant differences among versions. To evaluate identified contours, ground truth was defined as by radiation oncologists, indices for VA60 recalculated. Statistical analysis performed between each version. Among evaluated, nine did not satisfy established criteria. revealed brain, rectum, bladder differed substantially across AIRC In particular, pre-update rectum had (range) of 0.76 (0.40-1.16), whereas exhibited lower quality, 1.13 (0.24-5.68). Therefore, DLS methods undergo regular updates must be reassessed quality region interest. proposed help reduce workflows while appropriately evaluating performance.

Language: Английский

---

Trade‐off of different deep learning‐based auto‐segmentation approaches for treatment planning of pediatric craniospinal irradiation autocontouring of OARs for pediatric CSI

Medical Physics, Journal Year: 2025, Volume and Issue: unknown

Published: April 1, 2025

As auto-segmentation tools become integral to radiotherapy, more commercial products emerge. However, they may not always suit our needs. One notable example is the use of adult-trained software for contouring organs at risk (OARs) pediatric patients. This study aimed compare three approaches in context craniospinal irradiation (CSI): commercial, out-of-the-box, and in-house. CT scans from 142 patients undergoing CSI were obtained St. Jude Children's Research Hospital (training: 115; validation: 27). A test dataset comprising 16 was collected McGill University Health Centre. All images underwent manual delineation 18 OARs. LimbusAI v1.7 served as product, while nnU-Net trained benchmarking. Additionally, a two-step in-house approach pursued where smaller 3D containing OAR interest first recovered then used input train organ-specific models. Three variants U-Net architecture explored: basic U-Net, an attention 2.5D U-Net. The dice similarity coefficient (DSC) assessed segmentation accuracy, DSC trend with age investigated (Mann-Kendall test). radiation oncologist determined clinical acceptability all contours using five-point Likert scale. Differences between validation datasets reflected distinct institutional standards. lungs left kidney displayed increasing age-related values on datasets. esophagus often truncated distally mistaken trachea younger patients, resulting score less than 0.5 both kidneys frequently exhibited false negatives, leading mean that up 0.11 lower set 0.07 compared other Overall, achieved good performance body but difficulty differentiating laterality head structures, large variation standard deviation reaching 0.35 lenses. models generally had similar when against each nnU-Net. Inference time data 47-55 min Central Processing Unit (CPU) models, it 1h 21m V100 Graphics (GPU) could adapt well anatomy kidneys. When do population, viable option requires adjustments. In resource-constrained settings, model provides alternative. Implementing automated tool careful monitoring quality assurance regardless approach.

Language: Английский

---

A Japanese national survey on IMRT/SBRT in 2023 by the JASTRO High-Precision External Beam Radiotherapy Group

Journal of Radiation Research, Journal Year: 2025, Volume and Issue: unknown

Published: April 16, 2025

Abstract The purpose of this study was to investigate the utilization and implementation stereotactic body radiotherapy (SBRT) intensity-modulated (IMRT) in Japan up 2023. survey conducted by Japanese Society for Radiation Oncology High-Precision External Beam Radiotherapy Group Subcommittee from December 2023 February 2024. targeted patients treated with IMRT or SBRT between January 2021 2022. A comprehensive web-based questionnaire distributed 880 facilities, separate sections radiation oncologists medical physicists/radiotherapy technologists. total 360 facilities responded (response rate: 40.9%) section oncologists, 405 46.0%) technologists, providing data on status, techniques, workload challenges associated SBRT. Based responses used 68.6% responding institutes, 87.8%. VMAT emerged as most common technique (78.3%). highlighted a high demand physicists perform (86.9%). 84.6% that have not performed reported main reason lack oncologists. Furthermore, also noted significant variations prescribed doses margin sizes across indicating need further standardization. High-precision techniques such are getting popular, however, facility requirements which mandate presence at least two prevents becoming more widespread Japan.

Language: Английский

---

PURE-MRI: An International Study Assessing Physician Accuracy in Delineating the Prostate and Urethra on Prostate MRI

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2025, Volume and Issue: unknown

Published: April 25, 2025

Abstract Purpose Precise delineation of genitourinary structures during prostate cancer (PCa) care is critical to optimize treatment delivery while minimizing toxicity and injury. The Prostate UREthra on MRI (PURE-MRI) study was an international, prospective assess physicians’ accuracy segmenting urethra MRI. Methods Physicians who diagnose or treat PCa were invited contour patient cases using standard T 2 -weighted (all planes). We compared these contours reference consensus segmentations produced by a multidisciplinary panel experts. also evaluated performance validated auto- segmentation AI tool. Accuracy assessed with spatial volumetric analyses. Mixed effects model used evaluate potential factors influencing performance. Results 62 specialists from 11 countries created 114 110 contours. median (min, max) Dice score 0.92 (0.62, 0.95) for physicians. There no clear effect clinical experience focus. Maximum deviation inside (under-segmentation), maximum beyond expert contour, mean (per case) the 3.4 mm (1.0, 12.4), 5.3 (2.4, 17.3), 1.6 (0.9, 3.9), respectively. In comparison, auto-segmentation tool results 0.95 (0.94, 0.96), 3.0 mm, 3.9 (3.1, 4.9), 1.2 (1.1, 1.6), Physician considerably worse urethra, 0.33 (0.03, 0.69). No tested. Conclusion overall >0.9, though typically had errors >5 sometimes >10 mm. These patterns observed regardless experience, specialty, performs well enough use, given comparable practicing contrast, challenging. More training, better imaging, and/or tools may be necessary achieve consistent, accurate urethra.

Language: Английский

---

Quantifying and visualising uncertainty in deep learning-based segmentation for radiation therapy treatment planning: What do radiation oncologists and therapists want?

Radiotherapy and Oncology, Journal Year: 2024, Volume and Issue: 201, P. 110545 - 110545

Published: Sept. 24, 2024

Language: Английский

---

Geometric and dosimetric evaluation for breast and regional nodal auto‐segmentation structures

Journal of Applied Clinical Medical Physics, Journal Year: 2024, Volume and Issue: 25(10)

Published: Aug. 2, 2024

Abstract The accuracy of artificial intelligence (AI) generated contours for intact‐breast and post‐mastectomy radiotherapy plans was evaluated. Geometric dosimetric comparisons were performed between auto‐contours (ACs) manual‐contours (MCs) produced by physicians target structures. Breast regional nodal structures manually delineated on 66 breast cancer patients. ACs retrospectively generated. characteristics the breast/post‐mastectomy chestwall (CW) (axillary [AxN], supraclavicular [SC], internal mammary [IM]) geometrically evaluated Dice similarity coefficient (DSC), mean surface distance, Hausdorff Distance. also dosimetrically superimposing MC clinically delivered onto to assess impact utilizing with dose (Vx%) evaluation. Positive geometric correlations volume DSC intact‐breast, AxN, CW observed. Little or anti IM SC shown. For plans, insignificant differences MCs observed AxN V95% ( p = 0.17) 0.16), while IMN V90% significantly different. average (98.4%) (97.1%) comparable but statistically different 0.02). 0.35) 0.08) consistent MCs, Additionally, 94.1% AC‐breasts met ΔV95% variation <5% when > 0.7. However, only 62.5% AC‐CWs achieved same metrics, despite AC‐CW 0.43) being insignificant. AC structure similar MCs. may require manual adjustments. Careful review should be before treatment planning. findings this study guide clinical decision‐making process utilization AI‐driven plans. Before implementation auto‐segmentation software, an in‐depth assessment agreement each local facilities is needed.

Language: Английский

---