Division of Oral and Maxillofacial Radiology, Department of Oral Biology and Diagnostic Sciences, Faculty of Dentistry, Chiang Mai University, Suthep Road, Suthep Sub-district, Mueang Chiang Mai District, Chiang Mai 50200, Thailand.
Division of Computer Engineering, School of Information and Communication Technology, University of Phayao, Phahon Yothin Road, Mae Ka Sub-district, Mueang Phayao District, Phayao 56000, Thailand.
Int J Med Inform. 2025 Jan;193:105666. doi: 10.1016/j.ijmedinf.2024.105666. Epub 2024 Oct 28.
Radiolucent jaw lesions like ameloblastoma (AM), dentigerous cyst (DC), odontogenic keratocyst (OKC), and radicular cyst (RC) often share similar characteristics, making diagnosis challenging. In 2021, CrossViT, a novel deep learning approach using multi-scale vision transformers (ViT) with cross-attention, emerged for accurate image classification. Additionally, we introduced Extended Cropping and Padding (ECAP), a method to expand training data by iteratively cropping smaller images while preserving context. However, its application in dental radiographic classification remains unexplored. This study investigates the effectiveness of CrossViTs and ECAP against ResNets for classifying common radiolucent jaw lesions.
We conducted a retrospective study involving 208 prevalent radiolucent jaw lesions (49 AMs, 59 DCs, 48 OKCs, and 54 RCs) observed in panoramic radiographs or orthopantomograms (OPGs) with confirmed histological diagnoses. Three experienced oral radiologists provided annotations with consensus. We implemented horizontal flip and ECAP technique with CrossViT-15, -18, ResNet-50, -101, and -152. A four-fold cross-validation approach was employed. The models' performance assessed through accuracy, specificity, precision, recall (sensitivity), F1-score, and area under the receiver operating characteristics (AUCs) metrics.
Models using the ECAP technique generally achieved better results, with ResNet-152 showing a statistically significant increase in F1-score. CrossViT models consistently achieved higher accuracy, precision, recall, and F1-score compared to ResNet models, regardless of ECAP usage. CrossViT-18 achieved the best overall performance. While all models showed positive ability to differentiate lesions, DC had the highest AUCs (0.89-0.90) and OKC the lowest (0.72-0.81). Only CrossViT-15 achieved AUCs above 0.80 for all four lesion types.
ECAP, a targeted padding data technique, improves deep learning model performance for radiolucent jaw lesion classification. This context-preserving approach is beneficial for tasks requiring an understanding of the lesion's surroundings. Combined with CrossViT models, ECAP shows promise for accurate classification, particularly for rare lesions with limited data.
成釉细胞瘤（AM）、牙源性角化囊肿（OKC）、含牙囊肿（DC）和根侧囊肿（RC）等透光性颌骨病变具有相似的特征，这使得诊断具有挑战性。2021 年，CrossViT 作为一种新的深度学习方法，使用带有交叉注意力的多尺度视觉转换器（ViT），在图像分类方面取得了很高的准确率。此外，我们还引入了扩展裁剪和填充（ECAP），这是一种通过迭代裁剪较小的图像来扩展训练数据的方法，同时保留上下文。然而，它在牙科放射学分类中的应用尚未得到探索。本研究旨在调查 CrossViTs 和 ECAP 对分类常见透光性颌骨病变的 ResNets 的有效性。
我们进行了一项回顾性研究，纳入了 208 例经病理证实的常见透光性颌骨病变（49 例 AM、59 例 DC、48 例 OKC 和 54 例 RC），这些病变是在全景片或曲面体层片（OPG）中观察到的。三位有经验的口腔放射科医生进行了共识标注。我们使用 CrossViT-15、-18、ResNet-50、-101 和 -152 实现了水平翻转和 ECAP 技术。采用四折交叉验证方法。通过准确率、特异性、精度、召回率（敏感性）、F1 评分和接收者操作特征（ROC）曲线下面积（AUCs）等指标评估模型的性能。
使用 ECAP 技术的模型通常能获得更好的结果，其中 ResNet-152 的 F1 评分有显著提高。无论是否使用 ECAP，CrossViT 模型的准确率、精度、召回率和 F1 评分均高于 ResNet 模型。CrossViT-18 表现出最佳的整体性能。虽然所有模型都显示出对病变的区分能力，但 DC 的 AUC 值最高（0.89-0.90），OKC 的 AUC 值最低（0.72-0.81）。只有 CrossViT-15 对所有四种病变类型的 AUC 值均超过 0.80。
ECAP 是一种有针对性的填充数据技术，可提高透光性颌骨病变分类的深度学习模型性能。这种保留上下文的方法有利于需要理解病变周围环境的任务。CrossViT 模型结合 ECAP 有望实现准确分类，特别是对数据有限的罕见病变。
