S1574
Physics - Autosegmentation
ESTRO 2026
e85.[2] van der Velden, S., Simões, R., Gooding, M. J., Boukerroui, D., Remeijer, P., & Janssen, T. (2025). 2714 Assessing the long-term clinical usage of auto- segmentation for head and neck organs-at- risk. Radiotherapy and Oncology, 206, S2491-S2492. Keywords: autosegmentation,editing variability,acceptability Digital Poster 2772 Artificial intelligence in radiation oncology: Bridging gaps in cancer care in a low- and middle- income country Nowshin Taslima Hossain 1 , Aditi Paul Chowdhury 1 , Jannatul Ferdause 1 , A.M.M. Shariful Alam 1 , Masudul Hasan Arup 2,1 , Tasfi Jahan Tina 1 , Sura Jukrup Mumtahina 1 , Bhaskor Chakraborty 1 , Sumon Paul 1 , Md. Habibur Rahman 3 , Md. Abul Ahsan Didar 3 , Md. Raihan Bin Sharif 3 , Rajani Jha 4 , Sadia Sadiq 5 , Sweta Soni 6 , M. Saiful Huq 7 1 Radiation Oncology, Ahsania Mission Cancer and General Hospital, Dhaka, Bangladesh. 2 Radiation Oncology, National Institute of Cancer Research and Hospital, Dhaka, Bangladesh. 3 Medical Oncology, Ahsania Mission Cancer and General Hospital, Dhaka, Bangladesh. 4 Radiation Oncology, Kathmandu Cancer Center, Kathmundhu, Nepal. 5 Radiation Oncology, Inmol Hospital, Lahore, Pakistan. 6 Radiation Oncology, All India Institute of Medical Sciences, Rajkot, India. 7 Radiation Oncology, UPMC Hillman Cancer Center and University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, USA Purpose/Objective: Artificial Intelligence (AI) is increasingly utilized in radiation oncology to improve contouring accuracy and efficiency. In Bangladesh, approximately 167,526 new cancer cases are diagnosed annually, yet only 250 radiation oncologists and 39 radiotherapy machines are available nationwide. This mismatch results in treatment delays of up to 8 months in government hospitals and 1–2 months in private centers. At Ahsania Mission Cancer and General Hospital (AMCGH), nine radiation oncologists manage over 2,400 patients annually using two LINACs, one cobalt unit, and two contouring workstations. To address the heavy workload and limited manpower, AMCGH implemented the MVision AI-assisted auto-contouring system, the first such integration in Bangladesh, aiming to assess its feasibility, accuracy, and impact on workflow. Material/Methods: The MVision Contour Plus system was deployed and evaluated between August and October 2025. Total 91 cases were contoured using AI. Forty randomly selected cases were analyzed for organs at risk (OARs)
in the head and neck and pelvic regions. AI-generated contours were compared with manually delineated contours using Dice Similarity Coefficient (DSC), surface DSC (s-DSC), 95th percentile Hausdorff Distance (HD95), and volume difference metrics. Contouring times were recorded. Paired t-tests assessed time differences, while Wilcoxon signed-rank tests evaluated non-parametric DSC and HD95 comparisons between AI and manual contours. Results: For head and neck OARs (brainstem, eyes, lenses, optic chiasm), DSC values ranged 0.58–0.91, highest for eyes (0.90–0.91) and lowest for the optic chiasm (0.58); mean HD95 ranged 1.5–4.8 mm. Pelvic OARs (bladder, bowel bag, rectum, femoral heads, spinal cord) had a mean DSC ~0.80, highest for bladder and femoral heads (0.88–0.89) and lowest for spinal cord (0.55; HD95 = 50.4 mm). AI assisted contouring significantly reduced time from 25.4 ± 4.1 min to 7.6 ± 2.3 min (p < 0.001), a 70% reduction. No significant difference was observed in DSC for high-performing OARs (p > 0.05), while lower-performing structures (e.g., spinal cord, optic chiasm) showed expected variability. Inter-observer DSC (~0.80) was comparable to AI–manual results (p = 0.42). Radiation oncologists reported that AI’s integrated contouring guidelines were particularly valuable for complex OARs such as the hippocampus and brachial plexus. Conclusion: AI-assisted contouring is feasible, accurate, and time- efficient in high-volume, resource-constrained settings. By reducing contouring time by up to 70% (p < 0.001) without compromising clinical precision, AI integration can enhance workflow efficiency, consistency, and equitable access to timely radiotherapy in low- and middle-income countries, making it a practical necessity rather than a luxury. References: Huq MS, Acharya SC, Sapkota S, Silwal SR, Gautam M, Sharma S, Poudyal S, Sumon MA, Hossain T, Uddin AK, Gunasekara S. Cancer education and training within the South Asian Association for Regional Cooperation (SAARC) countries. The Lancet Oncology. 2024 Dec 1;25(12):e663-74.McGinnis GJ, Ning MS, Beadle BM, Joubert N, Shaw W, Trauernich C, Simonds H, Grover S, Cardenas CE, Court LE, Smith GL. Barriers and facilitators of implementing automated radiotherapy planning: A multisite survey of low-and middle-income country radiation oncology providers. JCO global oncology. 2022 May;8:e2100431. Keywords: Auto-contouring, Low-resource countries
Made with FlippingBook - Share PDF online