A fully automated sex estimation for proximal femur X-ray images through deep learning detection and classification

Li, Yuan; Niu, Chaoqun; Wang, Jian; Xu, Yong; Dai, Hao; Xiong, Tu; Yu, Dong; Guo, Huili; Liang, Weibo; Deng, Zhenhua; Lv, Jiancheng; Zhang, Lin

doi:10.1016/j.legalmed.2022.102056

Cited by 4 publications

(7 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The 35 studies show that AI applications are developed in thanatology, especially for postmortem identification [ 13 – 19 ], the estimation of the postmortem interval [ 20 – 22 ], and the determination of the causes of death [ 23 – 31 ]. In clinical forensic medicine, AI models are mainly designed for age estimation [ 15 , 32 – 44 ] and gender determination [ 15 – 17 , 45 ]. One AI model is aimed for the assessment and management of risk of violent reoffending among prisoners [ 46 ] and one for bruises dating [ 47 ].…”

Section: Resultsmentioning

confidence: 99%

“…The units of the performance metrics are not clear (years or months) 2 Oura et al 2021 [ 29 ] Determination of the causes of death Multilayer perceptron Testing accuracy and F1 range from 0.94 to 1, recall ranges from 0.89 to 1, precision from 0.92 to 1, and AUC from 0.99 to 1. Averaged test accuracy is 0.98 2 Garland et al 2021 [ 30 ] Determination of the causes of death Convolutional neural network Accuracy and F1 scores are equal to 1 2 Ibanez et al 2022 [ 31 ] Determination of the causes of death Convolutional neural network Recall, precision, and F1 score are respectively 0.93 ± 0.05, 0.89 ± 0.03, and 0.91 ± 0.04 2 Li et al 2022 [ 45 ] Gender determination for clinical forensic medicine purposes Convolutional neural network Average accuracy is 0.946 in Chinese Han population and 0.829 in White population 2 …”

Section: Resultsmentioning

confidence: 99%

“…In this case, the model performance may be overestimated due to model overfitting [ 49 ]. This issue, common when developing machine learning models, was explicitly handled in 11 reports by techniques based on model architecture [ 34 , 41 ] and parameters [ 38 , 40 , 44 , 45 ], input data [ 41 ], and validation steps [ 20 , 43 , 46 , 47 ]. In [ 34 ] and [ 41 ], the model architecture was modified to reduce overfitting by dropout regularization, that is to say removing nodes in a model by a given probability in order to simplify it.…”

Section: Discussionmentioning

confidence: 99%

“…This technique is known to reduce the generalization error of the model [ 50 ]. AI models were also developed by transfer learning, that is to say the use of a pre-trained model which is then adapted for a specific task, such as age estimation [ 38 , 44 ] or gender determination [ 45 ] of living individuals. In [ 40 ], the model parameters were frozen in part of the models along the training phase to prevent overfitting.…”

Section: Discussionmentioning

confidence: 99%

See 3 more Smart Citations

Artificial intelligence in the practice of forensic medicine: a scoping review

Tournois,

Trousset,

Hatsch

et al. 2023

Int J Legal Med

View full text Add to dashboard Cite

Forensic medicine is a thriving application field for artificial intelligence (AI). Indeed, AI applications intended to forensic pathologists or forensic physicians have emerged since the last decade. For example, AI models were developed to help estimate the biological age of migrants or human remains. However, the uses of AI applications by forensic pathologists or physicians and their levels of integration in medicolegal practices are not well described yet. Therefore, a scoping review was conducted on PubMed, ScienceDirect, and Scopus databases. This review included articles that mention any AI application used by forensic pathologists or physicians in practice or any AI model applied in one expertise field of the forensic pathologist or physician. Articles in other languages than English or French or dealing mainly with complementary analyses handled by experts who are not forensic pathologists or physicians or with AI to analyze data for research purposes in forensic medicine were excluded from this review. All the relevant information was retrieved in each article from a grid analysis derived and adapted from the TRIPOD checklist. This review included 35 articles and revealed that AI applications are developed in thanatology and in clinical forensic medicine. However, those applications seem to mainly remain in research and development stages. Indeed, the use of AI applications by forensic pathologists or physicians is not actual due to issues discussed in this article. Finally, the integration of AI in daily medicolegal practice involves not only forensic pathologists or physicians but also legal professionals.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

See 2 more Smart Citations

Artificial intelligence in the practice of forensic medicine: a scoping review

Tournois,

Trousset,

Hatsch

et al. 2023

Int J Legal Med

View full text Add to dashboard Cite

show abstract

“…The proposed method utilized a Generative Adversarial Network with Long Short-term Memory and a 3D-Convolutional Neural Network classification model, providing an efficient alternative to biopsy-based techniques. Another interesting application of deep learning in X-ray image analysis is presented by a study [31] that developed a fully automated deep learning pipeline using digital radiographs to detect the proximal femur region for accurate automated sex estimation. The model, based on convolutional neural networks, achieved an accuracy similar to that of current state-of-the-art mathematical functions using manually extracted features for the Chinese Han population samples, proving to be a reliable choice for human sex estimation.…”

Section: Introductionmentioning

confidence: 99%

A Novel Deep Transfer Learning-Based Approach for Automated Pes Planus Diagnosis Using X-ray Image

et al. 2023

View full text Add to dashboard Cite

Pes planus, colloquially known as flatfoot, is a deformity defined as the collapse, flattening or loss of the medial longitudinal arch of the foot. The first standard radiographic examination for diagnosing pes planus involves lateral and dorsoplantar weight-bearing radiographs. Recently, many artificial intelligence-based computer-aided diagnosis (CAD) systems and models have been developed for the detection of various diseases from radiological images. However, to the best of our knowledge, no model and system has been proposed in the literature for automated pes planus diagnosis using X-ray images. This study presents a novel deep learning-based model for automated pes planus diagnosis using X-ray images, a first in the literature. To perform this study, a new pes planus dataset consisting of weight-bearing X-ray images was collected and labeled by specialist radiologists. In the preprocessing stage, the number of X-ray images was augmented and then divided into 4 and 16 patches, respectively in a pyramidal fashion. Thus, a total of 21 images are obtained for each image, including 20 patches and one original image. These 21 images were then fed to the pre-trained MobileNetV2 and 21,000 features were extracted from the Logits layer. Among the extracted deep features, the most important 1312 features were selected using the proposed iterative ReliefF algorithm, and then classified with support vector machine (SVM). The proposed deep learning-based framework achieved 95.14% accuracy using 10-fold cross validation. The results demonstrate that our transfer learning-based model can be used as an auxiliary tool for diagnosing pes planus in clinical practice.

show abstract