Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{208320,
        author = {Dr Pranay Kumar Singh and Dr Shradha Sharma and Dr Zoya Hussain and Dr Sunidhi Kataria},
        title = {Real - Time Forensic Age & Sex Estimation Using AI on Panoramic Radiographs},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {4},
        pages = {1327-1336},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=208320},
        abstract = {—Background: Legal medicine, forensic identification, and disaster victim identification (DVI) all depend on accurate estimates of age and sex. Conventional radiographic and morphological techniques take a lot of time and are prone to fluctuation both within and between observers. New developments in artificial intelligence (AI), especially deep convolutional neural networks (CNNs), provide a real-time, automatic, and objective substitute. Because they provide a thorough picture of maxillo-mandibular growth, tooth mineralization, and skeletal architecture, digital panoramic radiographs (also known as orthopantomograms, or OPGs) are a perfect diagnostic modality for AI deployment. 
Objective: Using digitized panoramic radiographs from a wide range of demographics, this study developed, cross-verified, and validated an autonomous, dual-task deep learning framework for the simultaneous, real-time assessment of biological sex and chronological age. 
Materials and Methods: 12,500 high-resolution digital OPGs of patients ranging in age from 4.5 to 25.0 years were gathered from multicentric archives to provide a diverse dataset. The dataset was subjected to normalization, anatomical masking, and strict quality screening. An attention-gated ResNet-101 backbone was incorporated into a customized, multi-task deep convolutional neural network architecture that was designed to process single OPG inputs and split into two parallel computational pathways: a binary classification network for sex determination and a regression network for chronological age estimation. A strict double-check validation procedure was put in place to guarantee complete authenticity and exclude algorithmic bias. A 20% hold-out test set and an 80% training/validation set were created from the dataset. An external validation cohort (n = 2,500) from geographically separate universities not included in the training phase was used to do cross-verification. The network's regions of interest were mapped using Grad-CAM (Gradient-weighted Class Activation Mapping) visualizations, which offered forensic accountability and cryptographic explainability for the model's decision-making processes.
RESULTS: Real-time inference speeds of 42 milliseconds per input were attained by the optimized multi-task AI architecture. Over the whole hold-out test set, the model showed an overall Mean Absolute Error (MAE) of ± 0.38 years (95% Confidence Interval [CI]: 0.34–0.42 years) for chronological age prediction. The mixed-dentition sub-cohort showed the highest precision (MAE of ± 0.21 years). The system's overall accuracy for determining biological sex was 94.6% (Area Under the Receiver Operating Characteristic curve [AUC-ROC] = 0.978). With an accuracy of 93.8% for sex categorization and an MAE of ± 0.41 years, external cross-verification verified the stability of the model. Grad-CAM heatmaps confirmed that the network prioritized mandibular ramus flexure and gonial angle morphology for sex categorization while heavily relying on the mineralization condition of mandibular third molars and canine pulp-to-tooth ratios for age progression. 
CONCLUSION: This paper presents a highly accurate, authenticated, and cross-verified AI system that can provide real-time simultaneous age and sex estimation from panoramic radiographs. This methodology meets the stringent evidence requirements needed for legal testimony, forensic casework, and international humanitarian DVI operations by offering objective measurements in addition to explainable attention maps.},
        keywords = {},
        month = {September},
        }