论文网址：Deep learning for brain disorder diagnosis based on fMRI images - ScienceDirect
英文是纯手打的！论文原文的summarizing and paraphrasing。可能会出现难以避免的拼写错误和语法错误，若有发现欢迎评论指正！文章偏向于笔记，谨慎食用
目录
[1. 心得](#1. 心得)
[2. 论文逐段精读](#2. 论文逐段精读)
[2.1. Abstract](#2.1. Abstract)
[2.2. Introduction](#2.2. Introduction)
[2.3. The overview of deep learning methods](#2.3. The overview of deep learning methods)
[2.3.1. Artificial intelligence, machine learning and deep learning](#2.3.1. Artificial intelligence, machine learning and deep learning)
[2.3.2. Brief introduction to deep learning techniques in fMRI analysis](#2.3.2. Brief introduction to deep learning techniques in fMRI analysis)
[2.4. Deep learning in brain disorder diagnosis](#2.4. Deep learning in brain disorder diagnosis)
[2.4.1. Functional connectivity model based approaches](#2.4.1. Functional connectivity model based approaches)
[2.4.2. 2D/3D image processing perspective](#2.4.2. 2D/3D image processing perspective)
[2.4.3. fMRI images as a time series](#2.4.3. fMRI images as a time series)
[2.4.4. Joint spatial and temporal feature exploration](#2.4.4. Joint spatial and temporal feature exploration)
[2.4.5. Other deep learning models and training techniques](#2.4.5. Other deep learning models and training techniques)
[2.4.6. Summary](#2.4.6. Summary)
[2.5. Challenges and future outlook](#2.5. Challenges and future outlook)
[2.5.1. Discussions](#2.5.1. Discussions)
[2.5.2. Future outlook](#2.5.2. Future outlook)
[3. Reference](#3. Reference)
1. 心得
（1）少儿科普是吧？不推荐已经很熟悉这个领域的人看
（2）我怎么才刚开始就要看不下去了
（3）仅用图表示最基础的结构（如CNN或AE），但用过量的文字讲述一万个模型，是一种，让人，极为没有耐心的写作方式
（4）有一种站在未来看过去的感觉...但是吧...有些过去不是非要看的...
2. 论文逐段精读
2.1. Abstract
①fMRI can be regarded as image, time series and image series
2.2. Introduction
①Introducing fMRI itself→applications→pros and cons→compared with other imaging methods→introducing ML and AI
2.3. The overview of deep learning methods
2.3.1. Artificial intelligence, machine learning and deep learning
（1）Artificial intelligence
①介绍了...AI...
②The relationship between AI, ML and DL:
（2）Machine learning
①Categories of ML: supervised learning, unsupervised Learning, semi-supervised learning and reinforcement learning
（3）Deep learning
①...有出色的性能？？受益于计算能力？？fine
2.3.2. Brief introduction to deep learning techniques in fMRI analysis
①Common use DNN: CNN and RNN
②Workflow of deep learning:
（1）Convolutional neural networks
①介绍CNN？？？？卷积？？！激活函数？？
②图像分类的简化CNN网络？？？图？和上一张图有什么不一样？？
（2）Recurrent neural network
①Famous RNN: LSTM and GRU
（3）Auto encoder and decoder
①An example AE:
2.4. Deep learning in brain disorder diagnosis
2.4.1. Functional connectivity model based approaches
（1）Functional connectivity model construction
①Static and dynamic FC construction methods:
（2）FC based DL methods
①ML pipeline:
（3）Direct use of functional connectivity measures
①Directly classify FC matrix
②疯狂地用文字介绍了一堆方法
（4）End-to-end model for disease classification
①继续介绍一堆模型
（5）Connectivity matrices as an analogy to 2D images
①...Some researchers consider fMRI as 2D images and Conv it
（6）Comment on functional connectivity based methods
①There are some noises in raw fMRI data
2.4.2. 2D/3D image processing perspective
（1）4D fMRI to 2D images conversion
①列举了一堆4D转2D然后卷积的办法，从引用就能看出这些办法其实很老了，全在2020之前，而且这本来就有那么点不合理（虽然这篇是22写的啦...所以其实对于现在的我们参考性不是很强）
（2）3D neural network
①扒拉了一堆模型
（3）Challenges and opportunities
①For limited samples, the prior kownledge is feasible
2.4.3. fMRI images as a time series
①继续列
2.4.4. Joint spatial and temporal feature exploration
①Lists models with spatial and temporal methods
2.4.5. Other deep learning models and training techniques
（1）Graph CNN and its applications
（2）Generative models
（3）Transfer learning
2.4.6. Summary
①作者觉得以后深度学习会在临床诊断中大放异彩
2.5. Challenges and future outlook
2.5.1. Discussions
①Non of a DL model can contain all the tasks of disease diagnosis（确实，作者当时写的时候大模型还没有风靡，可能确实觉得，一个模型很难顾及到所有方面）
②Challenges: a) costs, b) interpretability
2.5.2. Future outlook
①"尽管近年来使用 fMRI 图像在脑部疾病诊断方面取得了巨大成功，但距离临床诊断要求还很远"，没错其实...2024年仍然还有一些距离，还需要一点重大突破...
②Researchers should mix all the data together, including electronic medical record, EEG, structural MRI image etc...确实，现在多模态还发展得不错
3. Reference
Yin, W., Li, L., & Wu, F. (2022) 'Deep learning for brain disorder diagnosis based on fMRI images', Neurocomputing, 469: 332-345. doi: https://doi.org/10.1016/j.neucom.2020.05.113