Performances of Depression Detection through Deep Learning-based Natural Language Processing to Mandarin Chinese Medical Records: Comparison between Civilian and Military Populations

10.4103/TPSY.TPSY_9_22

Tai-Yu Chen；Hsuan-Te Chu；Yueh-Ming Tai；Szu-Nian Yang

area under curve (AUC) ； artificial intelligence ； bidirectional decoder representations from transformers ； receiver operating characteristic (ROC)

台灣精神醫學

36卷1期（2022 / 03 / 01）

32 - 38

英文

Objective: A certain portion of patients with depression is under-diagnosed and has attracted the attention in the field of natural language processing (NLP). In this study, we intended to explore the feasibility of transferring unstructured textual records into a screening tool to early detect depression. Methods: We recruited 22,355 medical records in Mandarin traditional Chinese from the psychiatry emergency department of a military psychiatry center from 2004 to 2019. We preprocessed all the context of present illness histories as corpus and the presence of clinical diagnoses of depression as an outcome. A state-of-the-art NLP model was developed based on a pretrained bidirectional encoder representation from transformers (BERT) model along with several convolutional neural network (CNN) and trained by the training set (80% of original data) of total samples (BERT_(general)) and of civilian samples (BERT_(civilian)) and of military samples (BERT_(military)) independently. The receiver operating characteristic (ROC) and area under curve (AUC) of three trained models were compared for predicting depression for the test dataset (20% of original data) of general and specific samples. Results: The experimental results demonstrated excellent performance of BERT_(general) for general samples (AUC = 0.93, sensitivity = 0.817, specificity = 0.920 for optimal cut-off point) and civilian sample (AUC = 0.91, sensitivity = 0.851, specificity = 0.851 for optimal cut-off point). BERT_(general) showed a significant underperformance of for military samples (AUC = 0.79, sensitivity = 0.712, specificity = 0.732, p ＜ 0.05 for optimal cut-off point). That of BERT_(military) was slight higher (AUC = 0.82, sensitivity = 0.708, specificity = 0.786 for optimal cut-off point) for military samples. Conclusion: This study showed the feasibility of applying deep learning technique as a depression-detection assistant tool in Mandarin Chinese medical records. However, the subjects' specific situation, e.g., military status, is warranted for further investigation.

1. Ahmad, I,Pothuganti, K(2020).Analysis of different convolution neural network models to diagnose Alzheimer’s disease.Mater Today Proc,37,2800-2812.
2. Ariyarathne, G,De Silva, S,Dayarathna, S(2020).ADHD identification using convolutional neural network with seed-based approach for fMRI data.Proceedings of the 2020 9th International Conference on Software and Computer Applications,Chicago, Illinois, USA:
3. Colic, S,He, JC,Richardson, JD(2021).A machine learning approach to identification of self-harm and suicidal ideation among military and police veterans.J Mil Veteran Fam Health,8,e20210035-e20210043.
4. Core Team R(2013).Core Team R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing, 2013..
5. Dai, HJ,Su, CH,Lee, YQ,Das, S,Blacker, D,Smoller, JW(2020).Deep learning-based natural language processing for screening psychiatric patients.Front Psychiatry,11,533949.
6. Devlin, J,Chang, MW,Lee, K,Toutanova, K(2018).,未出版
7. Eichstaedt, JC,Smith, RJ,Merchant, RM(2018).Facebook language predicts depression in medical records.Proc Natl Acad Sci,115,11203-11208.
8. Greenberg, J,Tesfazion, AA,Robinson, CS(2012).Screening, diagnosis, and treatment of depression.Mil Med,177,60-66.
9. Kennedy, JE,Reid, MW,Lu, LH(2019).Validity of the CES-D for depression screening in military service members with a history of mild traumatic brain injury.Brain Inj,33,932-940.
10. Mitchell, AJ,Vaze, A,Rao, S(2009).Clinical diagnosis of depression in primary care: A meta-analysis.Lancet,374,609-619.
11. Peng, Y,Yan, S,Lu, Z(2019).,未出版
12. Seal, A,Bajpai, R,Agnihotri, J(2021).DeprNet: A deep convolution neural network framework for detecting depression using EEG.IEEE Trans Instrum Meas,70,1-13.
13. Sheu, YH,Magdamo, C,Miller, M,Das, S,Blacker, D,Smoller, JW(2021).,未出版
14. Skopp, NA,Holland, KM,Logan, JE(2019).Circumstances preceding suicide in US soldiers: A qualitative analysis of narrative data.Psychol Serv,16,302.
15. Su, D,Zhang, X,He, K(2021).Use of machine learning approach to predict depression in the elderly in China: A longitudinal study.J Affect Disord,282,289-298.
16. Tang, M,Gandhi, P,Kabir, MA,Zou, C,Blakey, J,Luo, X(2019).,未出版
17. Thériault, FL,Garber, BG,Momoli, F(2019).Mental health service utilization in depressed Canadian armed forces personnel.Can J Psychiatry,64,59-67.
18. Tsui, FR,Shi, L,Ruiz, V(2021).Natural language processing and machine learning of electronic health records for prediction of first-time suicide attempts.JAMIA Open,4,ooab011.
19. Van Rossum, G,Drake, FL(2000).Python Reference Manual.Scotts Valley, California, the USA:iUniverse Indiana.
20. Vaswani, A,Shazeer, N,Parmar, N(2017).Advances in neural information processing systems.Proc Mach Learn Res,30,5998-6008.
21. World Health Organization: Fact Sheet on Depression 2018. Geneva, Switzerland: World Health Organization, 2018.
22. Zhang, W,Liu, H,Silenzio, VM(2020).Machine learning models for the prediction of postpartum depression: Application and comparison based on a cohort study.JMIR Med Inform,8,e15516.

1. (2024).Performance of Artificial Intelligence Models (Bidirectional Encoder Representations from Transformers + TextCNN) in Detecting Eight Psychiatric Diagnoses from Unstructured Texts Chinese Electronic Medical Records.台灣精神醫學,38(3),120-127.
2. (2024).Performances of Large Language Models in Detecting Psychiatric Diagnoses from Chinese Electronic Medical Records: Comparisons between GPT-3.5, GPT-4, and GPT-4o.台灣精神醫學,38(3),134-141.