Research on Detection of Alzheimer Disease Based on Image Fusion Technology

doi:10.12068/j.issn.1005-3026.2025.20230338

Abstract

Abstract:

The plasma samples of Alzheimer disease（AD） patients are collected using Fourier transform infrared-attenuated total reflection （FTIR-ATR） spectroscopy technology. Based on the FTIR-ATR spectral data of the plasma membrane samples， the spectral data are encoded into two-dimensional images by utilizing the Gram angular field （GAF） and Markov transition field （MTF）. Meanwhile， a neural network model based on the deep residual networks and attention mechanism is combined to conduct the screening and classification research on Alzheimer disease. The experimental results show that the GAF-MTF-CNN model can effectively improve the accuracy of spectral feature extraction. Additionally， the method of combining two-dimensional data with deep learning has better classification accuracy compared with traditional classification methods. Encoding spectrum into images using GAF and MTF techniques， and combining them with an improved residual neural network， effectively enhances the generalization ability and diagnostic accuracy of AD screening models， optimizing the screening performance.

Key words: near-infrared spectrum, Alzheimer disease, Gramian angular field （GAF）, Markov transition field （MTF）, convolutional neural networks （CNN）

CLC Number:

TP 391

Zhi-gang LI, Ming-kai MU, De-an HU, Nan XIANG. Research on Detection of Alzheimer Disease Based on Image Fusion Technology[J]. Journal of Northeastern University(Natural Science), 2025, 46(6): 1-7.

Figures/Tables 11

Fig.1 Near-infrared spectrum of 179 AD samples

Fig.2 Converting spectra into images

Fig.3 MTF conversion process

Fig.4 Six different images

Table 1 Experimental sample set division

模型数据集	训练集数目		测试集数目
模型数据集	AD	健康	AD	健康
AD早期与健康对照组	69	85	73	86
AD中期与健康对照组	18	32	19	34

Fig.5 Structure of the residual network model

Fig.6 Overall process of Alzheimer disease screening

Fig.7 Classification accuracy of different two-dimensional image sizes

Fig.8 Classification accuracy of different neural networks

Table 2 Classification metrics for early AD with

方法	图像尺寸	特征数	灵敏度/%	特异性/%	准确率/%
方法	像素	特征数	灵敏度/%	特异性/%	准确率/%
RF-relieff	—	28	94.44	87.5	90
PCA-KNN	—	28	83.33	90.63	88
PLSDA	—	200	89.89	96.88	94
本文方法	128×128	—	94.44	93.75	94

Table 3 Classification metrics for medium AD with

方法	图像尺寸	特征数	灵敏度/%	特异性/%	准确率/%
方法	像素	特征数	灵敏度/%	特异性/%	准确率/%
RF-relieff	—	28	89.47	88.23	88.68
PCA-KNN	—	28	73.68	91.18	84.91
PLSDA	—	200	89.47	91.18	90.57
本文方法	128×128	—	94.74	91.18	92.45

References 21

[1]	Brown E E， Kuumar S， Raiji T K， et al. Anticipating and mitigating the impact of the COVID-19 pandemic on Alzheimer’s disease and related dementias［J］. American Journal of Geriatric Psychiatry， 2020， 28（7）： 712-721.
[2]	Sala A， Anderson D J， Breman P M， et al. Biofluid diagnostics by FTIR spectroscopy：a platform technology for cancer detection［J］. Cancer Letters， 2020， 477： 122-130.
[3]	Backhaus J， Mueller R， Formanski N， et al. Diagnosis of breast cancer with infrared spectroscopy from serum samples［J］. Vibrational Spectroscopy， 2010， 52（2）： 173-177.
[4]	Harris C， Despa M， Kelly K. Design and fabrication of a cross flow micro heat exchanger［J］. Journal of Microelectromechanical Systems， 2000， 9（4）： 502-508.
[5]	Ollesch J， Heinze M， Michael H. It’s in your blood： spectral biomarker candidates for urinary bladder cancer from automated FTIR spectroscopy［J］. Journal of Biophotonics， 2014， 7（3/4）： 210-221.
[6]	单鹏，吴缀，何年，等.基于ATR-FTIR光谱的γ-PGA发酵批次分类研究［J］. 东北大学学报（自然科学版）， 2022， 43（10）： 1376-1382.
	Shan Peng， Wu Zhui， He Nian， et al. Research on batch classification of γ-polyglutamic-acid fermentation based on ATR-FTIR spectroscopy［J］. Journal of Northeastern University（Natural Science）， 2022， 43（10）： 1376-1382.
[7]	Du Y， Xie F， Yin L F， et al. Breast cancer early detection by using Fourier-transform infrared spectroscopy combined with different classification algorithms［J］. Spectrochimica Acta Part A： Molecular and Biomolecular Spectroscopy， 2022， 283： 121715.
[8]	Ryzhikova E， Kazakov O， Halamkova L， et al. Raman spectroscopy of blood serum for Alzheimer’s disease diagnostics：specificity relative to other types of dementia［J］. Journal of Biophotonics， 2015， 8（7）： 584-596.
[9]	吕庆坤，黄高忠. 外周血拉曼光谱在阿尔茨海默病诊断中的应用［J］. 中华老年多器官疾病杂志， 2017， 16（4）： 313-316.
	Qing-kun Lyu， Huang Gao-zhong. Application of Raman spectroscopy of peripheral blood in the diagnosis of Alzheimer’s disease［J］. Chinese Journal of Geriatric Multiple-Organ Diseases in the Elderly， 2017， 16（4）： 313-316.
[10]	Stahlschmidt S R， Ulfenborg B， Synnergren J. Multimodal deep learning for biomedical data fusion： a review［J］. Briefings in Bioinformatics， 2022， 23（2）： 569.
[11]	Wang Z G， Oates T. Imaging time-series to improve classification and imputation［EB/OL］. （2015-06-01）［2023-10-15］. .
[12]	Wang Q Y， Pian F F， Wang M X， et al. Quantitative analysis of Raman spectra for glucose concentration in human blood using Gramian angular field and convolutional neural network［J］. Spectrochimica Acta Part A： Molecular and Biomolecular Spectroscopy， 2022， 275： 121189.
[13]	De Paula P O， da Silva C T B， de Faissol A R R， et al. Classification of image encoded SSVEP-based EEG signals using convolutional neural networks［J］. Expert Systems with Applications， 2023， 214（15）： 119096.
[14]	雷春丽，夏奔锋，薛林林，等. 基于MTF-CNN的滚动轴承故障诊断方法［J］. 振动与冲击， 2022， 41（9）： 151-158.
	Lei Chun-li， Xia Ben-feng， Xue Lin-lin， et al. Rolling bearing fault diagnosis method based on MTF-CNN［J］. Journal of Vibration and Shock， 2022， 41（9）： 151-158.
[15]	Sun W， Zhou J， Sun B T， et al. Markov transition field enhanced deep domain adaptation network for milling tool condition monitoring［J］. Micromachines， 2022， 13（6）： 873.
[16]	张珂，冯晓晗，郭玉荣，等. 图像分类的深度卷积神经网络模型综述［J］. 中国图象图形学报， 2021， 26（10）： 2305-2325.
	Zhang Ke， Feng Xiao-han， Guo Yu-rong， et al. Overview of deep convolutional neural networks for image classification［J］. Journal of Image and Graphics， 2021， 26（10）： 2305-2325.
[17]	王一宁，秦品乐，李传朋，等. 基于残差神经网络的图像超分辨率改进算法［J］. 计算机应用， 2018， 38（1）： 246-254.
	Wang Yi-ning， Qin Pin-le， Li Chuan-peng， et al. Improved algorithm of image super resolution based on residual neural network［J］. Journal of Computer Applications， 2018， 38（1）： 246-254.
[18]	车畅畅，王华伟，倪晓梅，等. 基于深度残差收缩网络的滚动轴承故障诊断［J］. 北京航空航天大学学报， 2021， 47（7）： 1399-1406.
	Che Chang-chang， Wang Hua-wei， Ni Xiao-mei， et al. Fault diagnosis of rolling bearings based on deep residual shrinkage network［J］. Journal of Beijing University of Aeronautics and Astronautics， 2021， 47（7）： 1399-1406.
[19]	魏颖，林子涵，齐林，等. 基于空间自注意力机制和深度特征重建的脑MR图像分割方法［J］. 东北大学学报（自然科学版）， 2023， 44（2）： 177-185.
	Wei Ying， Lin Zi-han， Qi Lin， et al. Brain MR image segmentation based on spatial self-attention mechanism and depth feature reconstruction［J］. Journal of Northeastern University（Natural Science）， 2023， 44（2）： 177-185.
[20]	Xu Z B， Lee C， Lyu Y Q， et al. Ensemble capsule network with an attention mechanism for the fault diagnosis of bearings from imbalanced data samples［J］. Sensors， 2022， 22（15）： 5543.
[21]	Yin Y Q， Li J C， Ling C J， et al. Fusing spectral and image information for characterization of black tea grade based on hyperspectral technology［J］. LWT， 2023， 185： 115150.