Abstract: Aiming at the problems that the corresponding features between different modes are easy to be fused and mislocated, the subjective empirical parameter adjustment of recognition model experts, and the high computational cost,  a self-optimizing dual-mode ("contrast enhanced T1 weighting " and "high resolution enhanced T2 weighting") multi-channel non-deep vestibular schwannioma recognition model was proposed. Firstly, a vestibular schwannoma recognition model was constructed to further explore the multimode image features of vestibular schwannoma and the complex nonlinear complementary information among the modes. Then, a model optimization strategy based on the global parallel sparrow search algorithm of game theory was designed to realize the adaptive optimization of key hyperparameters of the model, so that the model had a better recognition effect. Experimental results show that compared with the deep learning-based method, the proposed method reduces the number of parameters by 27.9% with an improvement of 4.19 percentage points in recognition accuracy, which verifies the effectiveness and adaptability of the proposed method.

Key words: vestibular schwannoma, multimode neural network, non-deep model, parallel acceleration, model self-optimization

摘要： 针对不同模态间对应特征极易融合错位、识别模型专家主观经验式调参且计算成本高等问题，提出自优化双模态（“对比增强T1加权”与“高分辨率T2加权”）多通路非深度前庭神经鞘瘤识别模型。首先，通过构建前庭神经鞘瘤识别模型进一步挖掘前庭神经鞘瘤病症多模态影像特征及模态间复杂的非线性互补信息；然后，设计基于博弈论全局并行麻雀搜索算法的模型优化策略，实现模型关键超参数的自适应寻优，使模型具有较优的识别效果。实验结果表明，相较于基于深度学习的方法，所提方法在识别准确率提升4.19个百分点的情况下参数量降低了27.9%，验证了所提方法的有效性和自适应性。

关键词: 前庭神经鞘瘤, 多模态神经网络, 非深度模型, 并行加速, 模型自优化