Few study report important insights into methamphetamine (MA) abstainers’ brain functional pattern. A better understanding of underlying functional mechanism in the brains of MA abstainers will help to explain abnormal behaviors. 42 male MA abstainers who were currently in long-term abstinence status (at least 14 months) and 32 normal males were recruited. All subjects underwent functional magnetic resonance imaging (fMRI) while responding to drug clues. In this study, a convolutional neural network (CNN) recognition model based on short-time Fourier transform (STFT) was proposed to identify MA abstainers and normal control group. STFT provided the time-localized frequency information and CNN was used to extract the structure features of the time-frequency spectrograms. The results showed that the classifier achieved satisfactory performance (98.9% accuracy) and could extract stabile brain voxels information. The highly discriminative power voxels mainly concentrated in the left inferior frontal gyrus of the orbit, the bilateral postcentral gyrus and the bilateral paracentral lobule. This study provides a new insight into the difference functional pattern between MA abstainers and normal, which elucidates the pathological mechanism of MA abstainers from time-frequency spectrograms integration viewpoint.

中文翻译：

---

使用基于短时傅立叶变换的卷积神经网络识别甲基苯丙胺戒断者

很少有研究报告对甲基苯丙胺 (MA) 戒酒者的大脑功能模式的重要见解。更好地了解 MA 戒酒者大脑中的潜在功能机制将有助于解释异常行为。招募了 42 名目前处于长期禁欲状态（至少 14 个月）的男性 MA 戒酒者和 32 名正常男性。所有受试者在对药物线索作出反应的同时接受了功能磁共振成像 (fMRI)。在这项研究中，提出了一种基于短时傅立叶变换（STFT）的卷积神经网络（CNN）识别模型来识别 MA 弃权者和正常对照组。STFT 提供时域频率信息，CNN 用于提取时频频谱图的结构特征。结果表明，该分类器取得了令人满意的性能（98. 9% 准确率）并且可以提取稳定的大脑体素信息。高分辨力体素主要集中在左眼眶额下回、双侧中央后回和双侧中央旁小叶。本研究为了解 MA 戒断者与正常人之间的差异功能模式提供了新的视角，从时频频谱整合的角度阐明了 MA 戒断者的病理机制。