Kazimierczak Wojciech, Kazimierczak Natalia, Kędziora Kamila, Szcześniak Marta, Serafin Zbigniew
Department of Radiology and Diagnostic Imaging, Collegium Medicum, Nicolaus Copernicus University in Torun, Jagiellońska 13-15, 85-067 Bydgoszcz, Poland.
Kazimierczak Private Medical Practice, Dworcowa 13/u6a, 85-009 Bydgoszcz, Poland.
J Clin Med. 2024 Jun 20;13(12):3605. doi: 10.3390/jcm13123605.
: This study evaluates the diagnostic accuracy of an AI-assisted tool in assessing the proximity of the mandibular canal (MC) to the root apices (RAs) of mandibular teeth using computed tomography (CT). : This study involved 57 patients aged 18-30 whose CT scans were analyzed by both AI and human experts. The primary aim was to measure the closest distance between the MC and RAs and to assess the AI tool's diagnostic performance. The results indicated significant variability in RA-MC distances, with third molars showing the smallest mean distances and first molars the greatest. Diagnostic accuracy metrics for the AI tool were assessed at three thresholds (0 mm, 0.5 mm, and 1 mm). : The AI demonstrated high specificity but generally low diagnostic accuracy, with the highest metrics at the 0.5 mm threshold with 40.91% sensitivity and 97.06% specificity. : This study underscores the limited potential of tested AI programs in reducing iatrogenic damage to the inferior alveolar nerve (IAN) during dental procedures. Significant differences in RA-MC distances between evaluated teeth were found.
本研究使用计算机断层扫描（CT）评估一种人工智能辅助工具在评估下颌管（MC）与下颌牙齿根尖（RA）接近程度方面的诊断准确性。本研究纳入了57名年龄在18至30岁之间的患者，其CT扫描图像由人工智能和人类专家进行分析。主要目的是测量MC与RA之间的最短距离，并评估人工智能工具的诊断性能。结果表明，RA-MC距离存在显著差异，第三磨牙的平均距离最小，第一磨牙的平均距离最大。在三个阈值（0毫米、0.5毫米和1毫米）下评估了人工智能工具的诊断准确性指标。人工智能表现出高特异性，但总体诊断准确性较低，在0.5毫米阈值时指标最高，灵敏度为40.91%，特异性为97.06%。本研究强调了所测试的人工智能程序在减少牙科手术中对下牙槽神经（IAN）的医源性损伤方面的潜力有限。在所评估的牙齿之间发现了RA-MC距离的显著差异。
