MindMap Gallery 2023 Medical AI Industry Research
Including medical AIGC application scenarios, AI drugs, and in-depth medical care. For example, drug research and development is a long process, and there is an irreconcilable contradiction between the speed of AI and the slowness of pharmaceuticals.
Edited at 2023-07-28 17:27:47This is a mind map about bacteria, and its main contents include: overview, morphology, types, structure, reproduction, distribution, application, and expansion. The summary is comprehensive and meticulous, suitable as review materials.
This is a mind map about plant asexual reproduction, and its main contents include: concept, spore reproduction, vegetative reproduction, tissue culture, and buds. The summary is comprehensive and meticulous, suitable as review materials.
This is a mind map about the reproductive development of animals, and its main contents include: insects, frogs, birds, sexual reproduction, and asexual reproduction. The summary is comprehensive and meticulous, suitable as review materials.
This is a mind map about bacteria, and its main contents include: overview, morphology, types, structure, reproduction, distribution, application, and expansion. The summary is comprehensive and meticulous, suitable as review materials.
This is a mind map about plant asexual reproduction, and its main contents include: concept, spore reproduction, vegetative reproduction, tissue culture, and buds. The summary is comprehensive and meticulous, suitable as review materials.
This is a mind map about the reproductive development of animals, and its main contents include: insects, frogs, birds, sexual reproduction, and asexual reproduction. The summary is comprehensive and meticulous, suitable as review materials.
Medical AI
Gödel's incompleteness theorem
Gödel
Turing test
Human thinking has limitations, the formal systems established by humans are incomplete, and human logical thinking is not necessarily superior to computers.
emerge
Two technical paths
Model-based computing capabilities, such as Deep Blue
Training based on neural network models, such as AlphaGo
The particularity of medical AI
Interpretability: The workflow is consistent with human understanding of medical problems, or the workflow can be understood by humans.
Both the FDA and NMPA have set high requirements for the explainability of AI medical care
AIGC medical scene
Discriminative and generative techniques
Discriminant (probe capital)
logistic regression
Widely used in medical data analysis, medical management, etc.
Support vector machineSVM
Google’s liver cancer screening algorithm uses SVM to extract pixel-level features to effectively identify tumor and non-tumor regions
Decision tree DT
Microsoft's innerEye uses DT to process image data to determine the patient's likelihood of illness and the type of disease.
random forest
IBM Watson uses a random forest algorithm to select high-confidence trajectory paths from a large number of patient records to assist doctors in formulating treatment plans.
Neural NetworkNN
GE Healthcare uses deep learning technology to improve diagnostic efficiency of medical imaging
Generative
Naive Bayes NB
Tempus Labs analyzes large amounts of patient data and genetic data to predict a patient's risk of cancer and tumors
K nearest neighbor KNN
IBM Watson uses technologies such as KNN to analyze patients' clinical data and health records to predict possible disease risks
Hidden Markov Model HMM
Luminex uses HMM to develop tools for protein interaction analysis and genotype analysis
Variational autoencoder VAE
Aidoc uses VAE to automatically generate reports on X-rays and CT scans
Generative Adversarial NetworkGAN
DeepMind partners with Moorfields Eye Hospital to develop medical image generator using GAN technology
Application scenarios
R&D, production, sales
drug discovery
Preclinical studies
clinical research
Approval for listing
Sale
Diagnosis and treatment link
Large-scale applications are mainly in the field of chat robots
Directions for improvement: Strengthen user privacy protection and introduce professional doctors for quality supervision
Industry chain sorting
Industrial chain
open ecology
AIDD
AIDD
historical
Capital boost
Stone from other mountains
The time has not come yet
three standards
The industry has accumulated rich human experience
Experience can be quantified and defined as algorithms
Experience must be graded to be able to judge whether it is good or bad.
Far away
Although massive amounts of data have been accumulated in the field of drug development, even more remains unknown.
We have some crude classification and quantification of some diseases, but quantitative measurements of health and phenotypes are just beginning.
There is some classification of pros and cons and positive and negative feedback that can be used to optimize experience, but it is far from enough and the feedback cycle is long.
worth looking forward to
There's a bright future
No medicine
In the early stages of development, technology faces challenges with data algorithms and talent, and its role in drug discovery is limited to speed and cost rather than the quality of decision-making.
Drug research and development is a long process, and there is an irreconcilable contradiction between the speed of AI and the slowness of pharmaceuticals.
AI for science
Protein engineering, AI for protein design
In disciplines such as biomedicine and materials engineering, there are too few features extracted by humans. The data extracted by AI general large models are higher dimensional, more accurate, and lower cost.
Depth medicine
cure
Baumol cost disease
"How to Flip the Healthcare Price Curve: Firing Doctors"
Hiring others to obtain services is becoming increasingly expensive
Various applications
Psychological consultation, medical record preservation, medical translation, medication management, disease monitoring
Ethics and data
Autopilot
Data silos
Medical big data
Ownership lies with the individual, control lies with the hospital, and management lies with the government
The company and the hospital sign a research and development agreement to obtain it. The data is not discharged, but the model is discharged.
Understanding data is harder than getting it