Automatic Segmentation of Periapical Radiograph Using Color Histogram and Machine Learning for Osteoporosis Detection

1. Title

Automatic Segmentation of Periapical Radiograph Using Color Histogram and Machine Learning for Osteoporosis Detection

DOI: https://doi.org/10.1155/2023/6662911

2. Authors

Rini Widyaningrum, Enny Itje Sela, Reza Pulungan, Anindita Septiarini

3. Summary

Automated trabecular bone segmentation method on digital periapical radiographs for osteoporosis screening. Steps include ROI acquisition, grayscale conversion, clustering via color histogram (K-means vs. Fuzzy C-means), feature extraction, and ML classification (Decision Tree, Naive Bayes, MLP). Best performance: K-means (K=10) + MLP. Ground truth: DEXA.

4. Key Words

Osteoporosis Detection; Periapical Radiograph; Trabecular Bone; Image Segmentation; K-means; Machine Learning; Multilayer Perceptron; DEXA

5. Extracted Data

5.1 Year

2023

5.2 Modality

Digital Periapical Radiography (PSP)

5.3 Dataset

102 ROIs (52 Non-osteoporosis, 50 Osteoporosis). Postmenopausal women.

5.4 Dataset Split

Train: 60 (30/30) Test: 42 (22/20)

5.5 Network Architecture

Segmentation: K-means (K=8,10,12,15) & Fuzzy C-means Classification: MLP (2 hidden layers), J48 Decision Tree, Naive Bayes

5.6 Metrics

Best Model (K-means + MLP Test Set): Accuracy: 90.48% Specificity: 90.90% Sensitivity: 90.00% Time: < 1 minute/image

5.7 Ground-truth - Professional Qty

2

5.8 Ground-truth - Supervisor Presence

Yes

5.9 Ground-truth - Experience Level

Specialists

5.10 Ground-truth - Expertise Area

Dentistry; Medical Radiology

5.11 Ground-truth - Tool or System

DEXA; MATLAB; DBSWin 4.5

5.12 ML Task

Image Segmentation; Binary Classification

###** 5.13 Project Objective** Fast and automated mandibular trabecular bone screening for osteoporosis using routine periapical radiographs.

6. Clinical Relevance

Provides dentists with an automated, fast (<1 min) screening tool during routine dental visits to identify systemic bone loss, facilitating early referral for medical diagnosis and fracture prevention.