| Summary: | Osteoporosis is a chronic disease affecting bone tissue that may lead to fractures from minor accidents. Roughly 20% of females and 6 % of males have osteoporosis after age 50, but the disease might be present at a younger age. Early diagnosis is challenging because the disease has no symptoms. Dental radiography is a frequent examination that might be useful for early osteoporosis screening at dental clinics. This thesis explores the utility of radiomorphometric indices manually measured on panoramic radiographs and the feasibility of fully automated radiomorphometric indices for osteoporosis screening in Norwegian males and females. The data from the seventh survey of the Tromsø study (Tromsø7) were used. Participants aged 40 and older were examined with dental panoramic radiographs and dual-energy x-ray absorptiometry at the femoral neck. Other demographic, health, and lifestyle data were collected in questionnaires. Mandibular cortical width and shape were assessed. Thin ( ≤ 3 mm) and severely eroded cortex could differentiate osteoporotic from non-osteoporotic females. Combining mandibular cortical width and shape with Fracture Risk Assessment (FRAX) score improved their diagnostic efficacy. T-score was the strongest predictor of mandibular cortical morphology among other factors in females. In males, the T-score was weakly associated with cortical shape, while the efficacy estimates for radiomorphometric indices were inconclusive. The reproducibility of the manually measured indices was suboptimal. Nevertheless, developing a fully automated algorithm for measuring MCW was feasible. Its first step, localization of mental foramen, was best performed by EfficientDet neural network with an accuracy of 79%. To conclude, radiomorphometric indices might be as useful as existing risk-factor-based tools for osteoporosis screening in females, and their combination with the FRAX score has superior diagnostic efficacy. Future extensive studies should further explore the performance of fully automated radiomorphometric ...
|
|---|