Prediction of MMSE Score Using Time-Resolved Near-Infrared Spectroscopy
Prediction of MMSE Score Using Time-Resolved Near-Infrared Spectroscopy
复制标题
DOI：
10.1007/978-3-319-91287-5_23
复制
发表时间：
2018-01-01
期刊：
影响因子：
--
通讯作者：
Sakatani, Kaoru
中科院分区：
文献类型：
--
作者：
Oyama, Katsunori;Hu, Lizhen;Sakatani, Kaoru
时间分辨近红外光谱（TRS）可以评估前额叶皮层血红蛋白（Hb）的基线浓度，这反映了休息时区域脑血流量和神经元活动。在之前的一项研究中，我们证明了TRS测量的氧基hb、脱氧hb、总hb和氧饱和度（SO2）的基线浓度与迷你精神状态检查（MMSE）得分相关。在本研究中，我们研究了静息时用TRS测量的Hb浓度是否可以预测具有各种认知功能的老年人的MMSE评分。共纳入202例受试者，其中男性87例，女性115例，年龄73.4±13岁。首先，采用MMSE评估认知功能，然后用TRS测量双侧前额叶皮层氧基hb、脱氧hb、总hb和SO2的基线浓度。然后，我们采用深度神经网络（DNN）来预测MMSE评分。从对比结果来看，通过留一交叉验证，DNN的准确率为91.5%。我们发现，不仅SO2的基线浓度，光程长度也有助于预测MMSE评分。这些结果表明，TRS与DNN是一种有用的认知障碍筛查试验。
Time-resolved near-infrared spectroscopy (TRS) enables assessment of baseline concentrations of hemoglobin (Hb) in the prefrontal cortex, which reflects regional cerebral blood flow and neuronal activity at rest. In a previous study, we demonstrated that baseline concentrations of oxy-Hb, deoxy-Hb, total-Hb, and oxygen saturation (SO2) measured by TRS were correlated with mini mental state examination (MMSE) scores. In the present study, we investigated whether Hb concentrations measured with TRS at rest can predict MMSE scores in aged people with various cognitive functions. A total of 202 subjects (87 males, 115 females, age 73.4 +/- 13 years) participated. First, MMSE was conducted to assess cognitive function, and then baseline concentrations of oxy-Hb, deoxy-Hb, total-Hb, and SO2 in the bilateral prefrontal cortex were measured by TRS. Then, we employed the deep neural network (DNN) to predict the MMSE score. From the comparison results, the DNN showed 91.5% accuracy by leave-one-out cross validation. We found that not only the baseline concentration of SO2 but also optical path lengths contributed to prediction of the MMSE score. These results suggest that TRS with the DNN is useful as a screening test for cognitive impairment.