Computer methods and programs in biomedicine · 2025
A novel multimodal diagnostic framework integrating hyperspectral imaging and deep learning for predicting RET gene mutations in medullary thyroid carcinoma.
СтатьяВалидационное исследованиеНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
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.
Yang Zhenpeng, Su Peng, Zhang Yuyang, Zheng Haitao, Deng Yupeng, Lin Xiangfeng, Ding Changyuan, Li Wei, Liang Weili, Lv Bin
Аннотация
BACKGROUND AND OBJECTIVE: Medullary thyroid carcinoma (MTC) is an aggressive malignancy driven predominantly by activating mutations in the RET proto-oncogene. Conventional genotyping using polymerase chain reaction (PCR) or next-generation sequencing (NGS) is often hampered by burdensome costs and prolonged turnaround times, hindering timely clinical decision-making.
METHODS: We developed a rapid, cost-effective, multimodal deep-learning framework to predict RET mutations from standard H&E-stained slides. Our approach leverages hyperspectral imaging and integrates a 1D-CNN-LSTM network for spectral analysis with a Swin Transformer for spatial feature extraction. A cross-modal attention mechanism effectively fuses these representations. The model was trained and validated on 82 MTC cases from Qilu Hospital and externally tested on independent cohorts from two additional centers (n = 60).
RESULTS: The proposed framework achieved an overall accuracy of 89.5 %, with a sensitivity of 90.2 % and specificity of 88.6 % for RET mutation classification. External validation confirmed robust generalizability, with performance surpassing single-modality benchmarks by 7.0-19.5 %.
CONCLUSIONS: This study presents a non-invasive and efficient alternative for predicting RET mutations in MTC, demonstrating the potential of hyperspectral imaging and integrated deep learning to advance precision oncology.
METHODS: We developed a rapid, cost-effective, multimodal deep-learning framework to predict RET mutations from standard H&E-stained slides. Our approach leverages hyperspectral imaging and integrates a 1D-CNN-LSTM network for spectral analysis with a Swin Transformer for spatial feature extraction. A cross-modal attention mechanism effectively fuses these representations. The model was trained and validated on 82 MTC cases from Qilu Hospital and externally tested on independent cohorts from two additional centers (n = 60).
RESULTS: The proposed framework achieved an overall accuracy of 89.5 %, with a sensitivity of 90.2 % and specificity of 88.6 % for RET mutation classification. External validation confirmed robust generalizability, with performance surpassing single-modality benchmarks by 7.0-19.5 %.
CONCLUSIONS: This study presents a non-invasive and efficient alternative for predicting RET mutations in MTC, demonstrating the potential of hyperspectral imaging and integrated deep learning to advance precision oncology.