start-ver=1.4 cd-journal=joma no-vol=17 cd-vols= no-issue=1 article-no= start-page=29 end-page= dt-received= dt-revised= dt-accepted= dt-pub-year=2024 dt-pub=20241225 dt-online= en-article= kn-article= en-subject= kn-subject= en-title= kn-title=The Three-Class Annotation Method Improves the AI Detection of Early-Stage Osteosarcoma on Plain Radiographs: A Novel Approach for Rare Cancer Diagnosis en-subtitle= kn-subtitle= en-abstract= kn-abstract=Background/Objectives: Developing high-performance artificial intelligence (AI) models for rare diseases is challenging owing to limited data availability. This study aimed to evaluate whether a novel three-class annotation method for preparing training data could enhance AI model performance in detecting osteosarcoma on plain radiographs compared to conventional single-class annotation. Methods: We developed two annotation methods for the same dataset of 468 osteosarcoma X-rays and 378 normal radiographs: a conventional single-class annotation (1C model) and a novel three-class annotation method (3C model) that separately labeled intramedullary, cortical, and extramedullary tumor components. Both models used identical U-Net-based architectures, differing only in their annotation approaches. Performance was evaluated using an independent validation dataset. Results: Although both models achieved high diagnostic accuracy (AUC: 0.99 vs. 0.98), the 3C model demonstrated superior operational characteristics. At a standardized cutoff value of 0.2, the 3C model maintained balanced performance (sensitivity: 93.28%, specificity: 92.21%), whereas the 1C model showed compromised specificity (83.58%) despite high sensitivity (98.88%). Notably, at the 25th percentile threshold, both models showed identical false-negative rates despite significantly different cutoff values (3C: 0.661 vs. 1C: 0.985), indicating the ability of the 3C model to maintain diagnostic accuracy at substantially lower thresholds. Conclusions: This study demonstrated that anatomically informed three-class annotation can enhance AI model performance for rare disease detection without requiring additional training data. The improved stability at lower thresholds suggests that thoughtful annotation strategies can optimize the AI model training, particularly in contexts where training data are limited. en-copyright= kn-copyright= en-aut-name=HaseiJoe en-aut-sei=Hasei en-aut-mei=Joe kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=1 ORCID= en-aut-name=NakaharaRyuichi en-aut-sei=Nakahara en-aut-mei=Ryuichi kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=2 ORCID= en-aut-name=OtsukaYujiro en-aut-sei=Otsuka en-aut-mei=Yujiro kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=3 ORCID= en-aut-name=NakamuraYusuke en-aut-sei=Nakamura en-aut-mei=Yusuke kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=4 ORCID= en-aut-name=IkutaKunihiro en-aut-sei=Ikuta en-aut-mei=Kunihiro kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=5 ORCID= en-aut-name=OsakiShuhei en-aut-sei=Osaki en-aut-mei=Shuhei kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=6 ORCID= en-aut-name=HironariTamiya en-aut-sei=Hironari en-aut-mei=Tamiya kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=7 ORCID= en-aut-name=MiwaShinji en-aut-sei=Miwa en-aut-mei=Shinji kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=8 ORCID= en-aut-name=OhshikaShusa en-aut-sei=Ohshika en-aut-mei=Shusa kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=9 ORCID= en-aut-name=NishimuraShunji en-aut-sei=Nishimura en-aut-mei=Shunji kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=10 ORCID= en-aut-name=KaharaNaoaki en-aut-sei=Kahara en-aut-mei=Naoaki kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=11 ORCID= en-aut-name=YoshidaAki en-aut-sei=Yoshida en-aut-mei=Aki kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=12 ORCID= en-aut-name=FujiwaraTomohiro en-aut-sei=Fujiwara en-aut-mei=Tomohiro kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=13 ORCID= en-aut-name=NakataEiji en-aut-sei=Nakata en-aut-mei=Eiji kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=14 ORCID= en-aut-name=KunisadaToshiyuki en-aut-sei=Kunisada en-aut-mei=Toshiyuki kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=15 ORCID= en-aut-name=OzakiToshifumi en-aut-sei=Ozaki en-aut-mei=Toshifumi kn-aut-name= kn-aut-sei= kn-aut-mei= aut-affil-num=16 ORCID= affil-num=1 en-affil=Department of Medical Information and Assistive Technology Development, Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama University kn-affil= affil-num=2 en-affil=Science of Functional Recovery and Reconstruction, Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama University kn-affil= affil-num=3 en-affil=Department of Radiology, Juntendo University School of Medicine kn-affil= affil-num=4 en-affil=Plusman LCC kn-affil= affil-num=5 en-affil=Department of Orthopedic Surgery, Graduate School of Medicine, Nagoya University kn-affil= affil-num=6 en-affil=Department of Musculoskeletal Oncology and Rehabilitation, National Cancer Center Hospital kn-affil= affil-num=7 en-affil=Department of Musculoskeletal Oncology Service, Osaka International Cancer Institute kn-affil= affil-num=8 en-affil=Department of Orthopedic Surgery, Kanazawa University Graduate School of Medical Sciences kn-affil= affil-num=9 en-affil=Department of Orthopaedic Surgery, Hirosaki University Graduate School of Medicine kn-affil= affil-num=10 en-affil=Department of Orthopaedic Surgery, Kindai University Hospital kn-affil= affil-num=11 en-affil=Department of Orthopedic Surgery, Mizushima Central Hospital kn-affil= affil-num=12 en-affil=Science of Functional Recovery and Reconstruction, Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama University kn-affil= affil-num=13 en-affil=Science of Functional Recovery and Reconstruction, Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama University kn-affil= affil-num=14 en-affil=Science of Functional Recovery and Reconstruction, Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama University kn-affil= affil-num=15 en-affil=Science of Functional Recovery and Reconstruction, Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama University kn-affil= affil-num=16 en-affil=Science of Functional Recovery and Reconstruction, Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama University kn-affil= en-keyword=osteosarcoma kn-keyword=osteosarcoma en-keyword=medical image annotation kn-keyword=medical image annotation en-keyword=anatomical annotation method kn-keyword=anatomical annotation method en-keyword=rare cancer kn-keyword=rare cancer END