巴金森氏病全球罹患率約為0.3%，巴金森氏病患者主要的臨床症狀為顫抖、肢體僵硬、動作遲緩以及姿勢異常等，每年每一百萬人中約只有兩個人死於此病，但是患者需要被長期的照護，造成照護者的經濟與時間的負擔，因此如何早期診斷巴金森氏病就是重要的議題。目前診斷及鑑別巴金森氏病仍有困難，主要依靠醫學影像以評估腦部特定區域如紋狀體的功能性以及結構性的變化。99mTC-TRODAT-1放射性示蹤劑使用於 Single-photon emission computed tomography (SPECT) 影像來顯示多巴胺轉運體於紋狀體的情況，並且使用Binding Ratio (BR) 來量化評估。然而，現行的BR計算方式耗時且依賴專業的放射師圈選紋狀體的區域。近年來深度學習的盛行，電腦慢慢開始應用於輔助或代為執行許多任務，像是辨識、分類或是預測等等。鑒於此，本研究使用U-net、BCDU-net與LSTM-Unet三個深度學習模型來自動圈選紋狀體的區域，以減少計算BR所需的時間與人力成本。三個深度學習模型皆屬於U-net架構及其與Long Short-Term Memory (LSTM) 結合的變種，U-net架構能結合不同複雜度特徵進行訓練，而LSTM能利用時序上的相關性提升準確率，兩者的結合使得CT影像中不同複雜度的特徵以及相鄰的切片之間的相關性能被考慮進去進而提升圈選的準確度。經過訓練，在腦部CT影像中，U-net、BCDU-net及LSTM-Unet在圈選紋狀體的準確率分別約54%、60%及70%，結合LSTM的LSTM-Unet及BCDU-net相較於原始U-net都有準確率的提升。自動圈選紋狀體的技術可以提供更多切面的紋狀體位置以改善現行BR的計算方式，而且相較於現行的方式增加22.7% 紋狀體區域及減少15.4%非紋狀體區域的圈選，平均減少0.013BR的絕對誤差，平均圈選一張CT切片僅需要0.015秒。

About 0.3% of the world suffer from Parkinson’s disease (PD), and the main clinical symptoms are tremor, rigidity, bradykinesia, abnormal posture, etc. Only about 2 out of every million people die of this disease each year. However, patients need to be cared for a long time, which causes the economic and time cost of family. Therefore, how to early diagnose Parkinson’s disease is an important issue. In clinical, it’s still difficult to diagnose the Parkinson’s disease, mainly assess functional and structural changes in specific regions of the brain such as the striatum. 99mTC-TRODAT-1 radioactive tracer is used on Single-photon emission computed tomography (SPECT) images to show the dopamine transporter in the striatum, and the binding ratio (BR) is used for quantitative evaluation. Moreover, the current BR calculation method is time-consuming and needs manual delineation of the striatum region on the computed tomography (CT) images. With the prevalence of deep learning in recent years, deep learning assists and even performs many tasks, such as identification, classification, or prediction. This study uses U-net, BCDU-net, and LSTM-Unet to automatically delineate the striatum regions to reduce the time and labor cost. The three models use U-net architecture and combine the Long Short-Term Memory (LSTM). The U-net architecture combines multiple image scales features and LSTM take past and next appearance into consideration. The Jaccard Index of striatum segmentation in these models are about 54%, 60% and 70%, respectively. The proposed method delineates 15.4% fewer the non-striatum regions and 22.7% more the striatum regions in ROI, and reduces 0.013 mean absolute error (MAE) of BR compared to current method, and it only needs 0.15 seconds to predict striatum region on one CT image.

1. Marsden, C.D., Parkinson's disease. J Neurol Neurosurg Psychiatry, 1994. 57(6): p. 672-81.
2. Poewe, W., et al., Parkinson disease. Nature reviews Disease primers, 2017. 3(1): p. 1-21.
3. Wang, H., et al., Global, regional, and national life expectancy, all-cause mortality, and cause-specific mortality for 249 causes of death, 1980–2015: a systematic analysis for the Global Burden of Disease Study 2015. The lancet, 2016. 388(10053): p. 1459-1544.
4. Agid, Y., Parkinson's disease: pathophysiology. Lancet, 1991. 337(8753): p. 1321-4.
5. Gelb, D.J., E. Oliver, and S. Gilman, Diagnostic criteria for Parkinson disease. Arch Neurol, 1999. 56(1): p. 33-9.
6. Huang, W.-S., et al., Usefulness of brain 99m Tc-TRODAT-1 SPET for the evaluation of Parkinson’s disease. European journal of nuclear medicine and molecular imaging, 2004. 31(2): p. 155-161.
7. Booij, J., et al., Imaging of the dopaminergic neurotransmission system using single-photon emission tomography and positron emission tomography in patients with parkinsonism. Eur J Nucl Med, 1999. 26(2): p. 171-82.
8. Kung, M.P., et al., [99mTc]TRODAT-1: a novel technetium-99m complex as a dopamine transporter imaging agent. Eur J Nucl Med, 1997. 24(4): p. 372-80.
9. Chen, Y.K., et al., The role of dopamine transporter imaging agent [99mTc]TRODAT-1 in hemi-parkinsonism rat brain. Nucl Med Biol, 2001. 28(8): p. 923-8.
10. Mozley, P.D., et al., Binding of [99mTc]TRODAT-1 to dopamine transporters in patients with Parkinson's disease and in healthy volunteers. J Nucl Med, 2000. 41(4): p. 584-9.
11. Huang, W.S., et al., Evaluation of early-stage Parkinson's disease with 99mTc-TRODAT-1 imaging. J Nucl Med, 2001. 42(9): p. 1303-8.
12. Kononenko, I., Machine learning for medical diagnosis: history, state of the art and perspective. Artificial Intelligence in Medicine, 2001. 23(1): p. 89-109.
13. Shen, D., G. Wu, and H.-I. Suk, Deep Learning in Medical Image Analysis. Annual Review of Biomedical Engineering, 2017. 19(1): p. 221-248.
14. Curiac, D.-I., et al. Bayesian network model for diagnosis of psychiatric diseases. IEEE.
15. Illán, I.A., et al., Computer aided diagnosis of Alzheimer’s disease using component based SVM. Applied Soft Computing, 2011. 11(2): p. 2376-2382.
16. Al Jarullah, A.A. Decision tree discovery for the diagnosis of type II diabetes. IEEE.
17. Kong, B., et al. Recognizing end-diastole and end-systole frames via deep temporal regression network. in International conference on medical image computing and computer-assisted intervention. 2016. Springer.
18. Xie, Y., et al., Spatial Clockwork Recurrent Neural Network for Muscle Perimysium Segmentation. Med Image Comput Comput Assist Interv, 2016. 9901: p. 185-193.
19. Akkus, Z., et al., Deep Learning for Brain MRI Segmentation: State of the Art and Future Directions. Journal of Digital Imaging, 2017. 30(4): p. 449-459.
20. Andermatt, S., S. Pezold, and P. Cattin, Multi-dimensional Gated Recurrent Units for the Segmentation of Biomedical 3D-Data. 2016, Springer International Publishing. p. 142-151.
21. Brosch, T., et al., Deep 3D Convolutional Encoder Networks With Shortcuts for Multiscale Feature Integration Applied to Multiple Sclerosis Lesion Segmentation. IEEE Transactions on Medical Imaging, 2016. 35(5): p. 1229-1239.
22. Litjens, G., et al., A survey on deep learning in medical image analysis. Medical Image Analysis, 2017. 42: p. 60-88.
23. Xu, J., et al., A Fully Automatic Framework for Parkinson’s Disease Diagnosis by Multi-Modality Images. Frontiers in Neuroscience, 2019. 13.
24. Liu, S., et al. Early diagnosis of Alzheimer's disease with deep learning. IEEE.
25. Xu, L., et al., Automated Whole-Body Bone Lesion Detection for Multiple Myeloma on (68)Ga-Pentixafor PET/CT Imaging Using Deep Learning Methods. Contrast Media Mol Imaging, 2018. 2018: p. 2391925.
26. Faro, A., et al., Basal ganglia activity measurement by automatic 3-D striatum segmentation in SPECT images. IEEE Transactions on Instrumentation and Measurement, 2011. 60(10): p. 3269-3280.
27. Fang, Y.H., et al., Fully Automated Quantification of the Striatal Uptake Ratio of [(99m)Tc]-TRODAT with SPECT Imaging: Evaluation of the Diagnostic Performance in Parkinson's Disease and the Temporal Regression of Striatal Tracer Uptake. Biomed Res Int, 2015. 2015: p. 461625.
28. Jiang, C.-F., et al., Segmentation of the Striatum from MR Brain Images to Calculate the-TRODAT-1 Binding Ratio in SPECT Images. Computational and mathematical methods in medicine, 2013. 2013.
29. Makowski, C., et al., Evaluating accuracy of striatal, pallidal, and thalamic segmentation methods: Comparing automated approaches to manual delineation. Neuroimage, 2018. 170: p. 182-198.
30. Suetens, P., Fundamentals of medical imaging. 2017: Cambridge university press.
31. Ronneberger, O., P. Fischer, and T. Brox. U-net: Convolutional networks for biomedical image segmentation. in International Conference on Medical image computing and computer-assisted intervention. 2015. Springer.
32. Çiçek, Ö., et al. 3D U-Net: learning dense volumetric segmentation from sparse annotation. in International conference on medical image computing and computer-assisted intervention. 2016. Springer.
33. Zhou, Z., et al., Unet++: A nested u-net architecture for medical image segmentation, in Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support. 2018, Springer. p. 3-11.
34. Arbelle, A. and T.R. Raviv. Microscopy cell segmentation via convolutional LSTM networks. in 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019). 2019. IEEE.
35. Azad, R., et al. Bi-Directional ConvLSTM U-Net with Densley Connected Convolutions. in Proceedings of the IEEE International Conference on Computer Vision Workshops. 2019.
36. Hochreiter, S. and J. Schmidhuber, Long short-term memory. Neural computation, 1997. 9(8): p. 1735-1780.
37. Xingjian, S., et al. Convolutional LSTM network: A machine learning approach for precipitation nowcasting. in Advances in neural information processing systems. 2015.
38. Song, H., et al. Pyramid dilated deeper convlstm for video salient object detection. in Proceedings of the European conference on computer vision (ECCV). 2018.
39. Huang, G., et al. Densely connected convolutional networks. in Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.
40. Shiiba, T., et al., Improvement of classification performance of Parkinson’s disease using shape features for machine learning on dopamine transporter single photon emission computed tomography. PloS one, 2020. 15(1): p. e0228289.
41. Ferrante, E. and N. Paragios, Slice-to-volume medical image registration: A survey. Medical image analysis, 2017. 39: p. 101-123.
42. Li, G., et al., 3D and 4D medical image registration combined with image segmentation and visualization, in Clinical Technologies: Concepts, Methodologies, Tools and Applications. 2011, IGI Global. p. 885-894.
43. Mani, V. and S. Arivazhagan, Survey of medical image registration. Journal of Biomedical Engineering and Technology, 2013. 1(2): p. 8-25.
44. Dong, M., et al. Instance Segmentation from Volumetric Biomedical Images Without Voxel-Wise Labeling. in International Conference on Medical Image Computing and Computer-Assisted Intervention. 2019. Springer.
45. Vuola, A.O., S.U. Akram, and J. Kannala. Mask-RCNN and U-net ensembled for nuclei segmentation. in 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019). 2019. IEEE.
46. He, K., et al. Mask r-cnn. in Proceedings of the IEEE international conference on computer vision. 2017.
47. Guo, Z., et al., Deep learning-based image segmentation on multimodal medical imaging. IEEE Transactions on Radiation and Plasma Medical Sciences, 2019. 3(2): p. 162-169.
48. Hsu, S.-Y., et al., Feasible Classified Models for Parkinson Disease from 99mTc-TRODAT-1 SPECT Imaging. Sensors, 2019. 19(7): p. 1740.
49. Mabrouk, R., B. Chikhaoui, and L. Bentabet, Machine Learning Models Classification using Clinical and DaTSCAN SPECT Imaging features: A Study on SWEDD and Parkinson’s disease. Age. 61(11.2): p. 61-9.7.
50. Cheng, D. and M. Liu. CNNs based multi-modality classification for AD diagnosis. in 2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI). 2017. IEEE.
51. Cheng, D. and M. Liu. Combining convolutional and recurrent neural networks for Alzheimer's disease diagnosis using PET images. in 2017 IEEE International Conference on Imaging Systems and Techniques (IST). 2017. IEEE.
52. Leung, K., et al., A deep learning-based approach for disease detection in the projection space of DAT-SPECT images of patients with Parkinson’s disease. Journal of Nuclear Medicine, 2020. 61(supplement 1): p. 509-509.