การพัฒนาต้นแบบปัญญาประดิษฐ์ในการจำแนกเพศจากภาพถ่ายเอกซเรย์คอมพิวเตอร์ของกะโหลกด้านข้างบริเวณกระดูกสันคิ้วและสันกระดูกท้ายทอย ในประชากรไทย

Sunisa Aobaom; Supawit  Kanokthanasomboon; Aphinphon  Saensuk; Chanyanut  Rapatwongsathorn; Kochakorn  Phantawong

Authors

Sunisa Aobaom Department of Medical Technology, Faculty of Allied Health Sciences, Thammasat University Rangsit Campus, 99 Moo 18 Phahonyothin Road, Khlong Nueng, Khlong Luang, Pathum Thani 12121, Thailand. https://orcid.org/0000-0001-5660-3407
Supawit Kanokthanasomboon Undergraduate Program in Medical Technology, Faculty of Allied Health Sciences, Thammasat University
Aphinphon Saensuk Undergraduate Program in Medical Technology, Faculty of Allied Health Sciences, Thammasat University
Chanyanut Rapatwongsathorn Graduate Program in Forensic Science, Faculty of Allied Health Sciences, Thammasat University
Kochakorn Phantawong Department of Radiological Technology, Faculty of Allied Health Sciences, Thammasat University

Keywords:

Artificial Intelligence, Sex Classification, CT image, Supraorbital Ridge, Nuchal crest

Abstract

Background: The study of the human skull plays a crucial role in forensic science, particularly in cases where only partial remains, such as cranial fragments, are recovered. Challenges often arise from environmental conditions, time constraints, and religious considerations. Recently, the application of artificial intelligence (AI), especially deep learning algorithms, has shown promise in supporting sex estimation from cranial morphology with high accuracy.

Objective: This study aimed to develop a prototype AI model for sex classification using lateral cranial computed tomography (CT) images, with a focus on two anatomical landmarks: the supraorbital ridge and the nuchal crest, specifically within the Thai population.

Material and methods: A total of 190 lateral skull CT images were collected, divided into 150 training samples and 40 test samples. The Roboflow 3.0 Object Detection (Accurate) model, a deep Learning architecture (YOLOv8-compatible) was used to train two AI models: one with data augmentation and one without. Model performance was evaluated using mAP@50, precision, and recall metrics. In addition, results were compared with sex estimation accuracy derived from conventional anthropometric parameters: Maximum Cranial Length (MCL) and Lambda-opisthion Chord (OCC).

Results: The augmented model achieved a higher precision (98.1%) and recall (99.9%) compared to the non-augmented model, while both models yielded an identical mAP@50 of 99.5%. In the unknown dataset, the augmented model successfully detected 39 out of 40 cases, with high confidence scores, particularly in male subjects (up to 95%). In contrast, traditional anthropometric methods based on MCL and OCC measurements yielded lower accuracy at 76.31% and 52.65%, respectively.

Conclusion: The AI model developed in this study demonstrated high accuracy and stability in sex estimation from lateral skull images, especially when enhanced by data augmentation. This approach shows significant potential for application in forensic science, particularly within Thai forensic contexts where limited biological samples and time constraints are often encountered.

References

บางกอกบรอดคาสติ้ง แอนด์ ทีวี. ล่าเบาะแส ปริศนากะโหลกสาวผมแดง [อินเทอร์เน็ต]. 28 พฤษภาคม2568 [เข้าถึงเมื่อ 1 กรกฎาคม พ.ศ. 2568]. เข้าถึงได้จาก: https://news.ch7.com/detail/805391.

สยามรัฐ. เจ้าหน้าที่นิติเวชตรวจสอบชิ้นส่วนกะโหลกศีรษะและชิ้นส่วนสะบักซ้าย [อินเทอร์เน็ต]. 23 เมษายน 2567 [เข้าถึงเมื่อ 1 กรกฎาคม 2568]. เข้าถึงได้จาก: https://siamrath.co.th/n/531058

ประชาไทย. เช็คข้อพิพาทระหว่างกระบวนการชันสูตรพลิกศพกับหลักศาสนบัญญัติ [อินเทอร์เน็ต]. 1 มกราคม 2565 [เข้าถึงเมื่อ 1 กรกฎาคม 2568]. เข้าถึงได้จาก: https://prachatai.com/journal/2022/01/96693

Uabundit N, Chaiyamoon A, Iamsaard S, Yurasakpong L, Nantasenamat C, Suwannakhan A, Phunchago N. Classification and Morphometric Features of Pterion in Thai Population with Potential Sex Prediction. Medicina. 2021; 57(11):1282. doi: 10.3390/medicina57111282

Mahakkanukrauh P, Sinthubua A, Prasitwattanaseree S, Ruengdit S, Singsuwan P, Praneatpolgrang S, Duangto P. Craniometric study for sex determination in a Thai population. Anat Cell Biol. 2015;48(4):275–283. doi:10.5115/acb.2015.48.4.

Zhan MJ, Cui JH, Zhang K, Chen YJ, Deng ZH. Estimation of stature and sex from skull measurements by multidetector computed tomography in Chinese. Leg Med (Tokyo). 2019; 41:101625. doi:10.1016/j.legalmed.2019

Gallagher J. Announcing Roboflow Train 3.0. Roboflow Blog [Internet]. 2023 Jul. 11 [cited 2025 Nov. 9]; Available from:https://blog.roboflow.com/roboflow-train-3-0/

Packirisamy V, Aljarrah K, Nayak SB. Morphometric evaluation of the orbital region for sex determination in a Saudi Arabian population using 3DCT images. Anat Sci Int. 2024 ;99(1):118-126. doi: 10.1007/s12565-023-00742-6

Hoshioka Y, Torimitsu S, Makino Y, Yajima D, Chiba F, Yamaguchi R, Inokuchi G, Motomura A, Tsuneya S, Iwase H. Sex estimation from skull measurements of a contemporary Japanese population using three-dimensional computed tomography images. Int J Legal Med. 2025 ;139(1):383-391. doi: 10.1007/s00414-024-03319-8

Torimitsu S, Nakazawa A, Flavel A, Swift L, Makino Y, Iwase H, et al. Estimation of ancestry from cranial measurements based on MDCT data acquired in a Japanese and Western Australian population. Int J Legal Med. 2024;138:1193–1203. doi:10.1007/ s00414-024-03159-6

Simmons-Ehrhardt TL, Parks CL, Monson KL. Cranial and facial inter-landmark distances and tissue depth dataset from computed tomography scans of 388 living persons. Data Brief. 2022; 43:108334. doi:10.1016/j.dib. 2022.108334

Shorten C, Khoshgoftaar TM. A survey on Image Data Augmentation for Deep Learning. J Big Data. 2019 (60). doi: 10.1186/s40537-019-0197-0.

Elgendi M, Nasir MU, Tang Q, Smith D, Grenier JP, Batte C, et al. The effectiveness of image augmentation in deep learning networks for detecting COVID-19: a geometric transformation perspective. Front Med (Lausanne). 2021;8:629134. doi:10.3389/fmed. 2021.629134

RSD. Sex estimation of Brazilian skulls using discriminant analysis of cranial measurements. [Internet]. 2021 Aug. 10 [cited 2025 Jul. 30];10(10):e266101018760. Available from: https://rsdjournal.org/rsd/article/view/18760

Toy S, Secgin Y, Oner Z, Turan MK, Oner S, Senol D. A study on sex estimation by using machine learning algorithms with parameters obtained from computerized tomography images of the cranium. Sci Rep. 2022;12(1):4278. doi:10.1038/ s41598-022-07415-w

Lopez-Capp TT, Rynn C, Wilkinson C, Paiva LAS, Michel-Crosato E, Biazevic MGH. Sexing the Cranium from the Foramen Magnum Using Discriminant Analysis in a Brazilian Sample. Braz Dent J. 2018; 29(6):592-598. doi: 10.1590/0103-6440201802087.

Toneva D, Nikolova S, Agre G, Harizanov S, Fileva N, Milenov G, Zlatareva D. Enhancing Sex Estimation Accuracy with Cranial Angle Measurements and Machine Learning. Biology. 2024; 13(10):780. doi: 10.3390/biology13100780