DOI 10.1016/j.jormas.2026.102754
Evaluation of radiomics-based machine learning models for detection of cervical lymph node metastasis from oral cancer on PET images.
Аннотация
PURPOSE: In this study, we hypothesized that radiomics analysis applied to pre-treatment PET images would be more accurate in the detection of lymph node metastasis from oral squamous cell carcinoma than conventional methods based on SUVmax, SUVmean, MTV, and TLG.
MATERIALS AND
METHODS
Lymph nodes in which metastasis/non-metastasis was histopathologically confirmed after neck dissection for clinical diagnosis of oral cancer lymph node metastasis between 2016 and 2023 at a single institution were included in this retrospective study. Among these lymph nodes, those confirmed on pre-treatment PET imaging were subjected to lymph node segmentation on the PET imaging, and radiomics features were extracted from the resulting regions. We selected extracted features that were useful for detecting metastatic lymph nodes, built machine learning (ML) models, and verified their detection accuracy. For SUVmax, SUVmean, MTV, and TLG, the cutoff values for detecting metastatic lymph nodes and their accuracy were determined.
RESULTS
Sixty-seven subjects were finally included, and radiomics features were extracted from 129 lymph nodes (96 metastatic, 33 non-metastatic). The AUCs of the ML models constructed from the selected radiomics features ranged from 0.84 to 0.86. In the conventional method, the AUCs of SUVmax, SUVmean, MTV, and TLG were 0.82, 0.80, 0.71, and 0.73, respectively. In addition, the sensitivity of the ML models ranged from 84.8 % to 96.3 %, showing higher values than the 53.1 % to 79.2 % obtained by conventional methods.
CONCLUSIONS
Comparisons with conventional methods suggested that radiomics may be useful for the detection of metastatic lymph nodes.
Предметные рубрики MeSH
Подготовка и перевод публикации
Обработка статьи · этапы, ошибки и управление
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Тип публикации указан по данным источника. Уровень доказательности автоматически не присваивается.