AI is Changing Healthcare in Multiple Ways

AI is Changing Healthcare in Multiple Ways

Since OpenAI ’s official launch of their new conversational AI model, #chatgpt in November 2022, it has gone and taken the world by storm. ChatGPT has opened up avenues for exploration within multiple industries, and #healthcare and #biopharma is no exception.?

Forbes recently released an article on how ChatGPT is going to change healthcare forever . And they’re not wrong. Medicals researchers are already investigating all the ways Generative AI can transform everyday tasks. A quick search for ‘ChatGPT’ in The Lancet Digital Health reveals the early exploration of ChatGPT’s healthcare prospects, from supporting with writing patient clinic letters to medical publishing, and this is just the beginning.?

Although the introduction of ChatGPT and brought about abuzz with its potential use in the world of healthcare and biopharma, the intersection of AI & healthcare is not a new or recent concept and in fact, dates back to the 1970s. Since then, AI approaches have rapidly expanded and transformed the industries, by increasing efficiencies, reducing expenditure and ultimately delivering more effective patient outcomes.?

Artificial intelligence in lung cancer: current applications and perspectives, SpringerLink. ?

A little history - AI’s origins in healthcare?

Between the 1960s and 1970s, researchers began exploring #AI’s role in medical diagnosis and decision making. This produced the basis for the four different AI systems, one of which included #MYCIN - a rule-based AI system for treatment assistance, which recommended antibiotic treatment options based on patient case history.?

Pretty impressive for technology in 1976, right??

Although these early models did not lead to widespread use due to their limitations, they did go on to inspire the next generations of AI applications.?

From the 1990s, AI algorithms have been integrated in medical imaging analysis, detecting abnormalities and aiding with diagnosis. Newer AI techniques followed, such as machine learning and deep learning, to enhance drug discovery. In more recent years, AI has transformed personalized medicine developments based on patient data. One of the most recent milestones in the history of AI in healthcare, has been the use of AI during the COVID-19 pandemic , from screening, diagnostics, to prediction and results, showing AI’s capabilities to help combat a worldwide pandemic.?

It’s definitely undeniable that the effects of AI on healthcare delivery have been profound, from drug discovery to precision medicine, and the technology is only continuing to advance.?

#computervision and #diagnostics ?

The world of medical imaging and diagnostics has been transformed by a particular form of AI - computer vision. Computer vision uses algorithms, known as #neuralnetworks , to mimic human visual recognition capabilities. The algorithms are trained using large datasets, enabling the analysis of images and the recognition of complex patterns in order to form insight-based conclusions.?

By reducing the number of diagnostic errors, detecting anomalies that may be overlooked and analysing huge inflows of medical data, computer vision supports physicians with time optimisation, increasing efficiency and providing second opinions on diagnostic conclusions.?

Computer vision has become extremely successful and even life-saving, through its implementation in diagnostics and early disease detection in numerous medical disciplines.?

Use case of computer vision: Abnormality detection in radiology?

#radiology was one of the first medical disciplines to adopt computer vision-based applications. Computer vision can help analyse medical images, including chest X-rays and CT scans, in the detection of lung cancer. These algorithms are used to identify tumours and abnormalities, helping to make more accurate and faster diagnoses.?

No alt text provided for this image


Microsoft’s InnerEye is addressing a specific radiology challenge, which is the time-consuming task of analysing CT scans and manually segmenting anatomical structures. Using computer vision and machine learning, InnerEye assists clinicians to develop their own models for CT images, which is applicable to surgery planning and monitoring tumour progression. Using AI this way is accelerating clinician’s ability to perform radiotherapy planning 13 times faster. ?

?End-to-end image segmentation in radiotherapy using Microsoft InnerEye’s algorithm predictions.?

No alt text provided for this image


#syntheticdata and #clinicaltrials acceleration?

Biopharma is increasingly leveraging AI in clinical trials to accelerate timelines, enrol a more diverse pool of patients, streamline operational processes and enhance data analysis throughout the trial lifecycle.?

The development of new drugs is a lengthy and difficult process, and biopharma also faces patient-related challenges when collecting clinical trial data. One of these challenges is enrolling diverse populations in clinical trials, due to accessibility, recruitment and participation issues, leading to underrepresentation.?

However, synthetic data enables biopharma to diversify datasets and enhance clinical research by creating representative datasets, while also protecting sensitive patient information.?

Synthetic data is artificially generated information, trained on real world data samples, which can fill in missing data by producing entirely fabricated patient datasets for AI training.?

With more diverse trial groups, there is more data available on the traits associated with certain ethnic groups. This data can be used to develop synthetic patient models that represent more diverse populations and reduce bias in clinical trial studies.?

No alt text provided for this image
AI in drug development process

Since synthetic datasets are not connected to real individuals, this allows biopharma and CROs to focus on the data itself and speed up the early phase clinical trial process, rather than spend effort and resource on privacy management.?

Companies such as Medidata Solutions and Unlearn.AI are doing tremendous work deploying this technology into real life situations, supported by cloud platform technology firms like Amazon Web Services (AWS) and others.

Use case: Synthetic data and clinical trials?

Some companies are using synthetic control arms, instead of recruiting new patients, including the virtual reality company, AppliedVR . AppliedVR is conducting trials for their VR treatment to treat chronic back pain patients , by pulling data from Komodo Health’s existing database of chronic pain patients, as a comparison group. The synthetic data control arm consists of de-identified patient cohorts, similar to patients enrolled in the trial but are not included in the study itself. The database used also has information on patient race and ethnicity, so research teams can ensure minority groups are also represented in the trials.?

Although this approach to clinical trial design is not regularly used yet, using synthetic control groups can speed up the development of new therapies in a more representative and cost-effective way.?

Generative AI & #scientificcommunication ?

Aside from clinical and medical research, AI shows tremendous potential in scientific communication, particularly with the rise of generative AI and ChatGPT.?

Generative AI uses neural networks to learn how to generate new, unique and original outputs. Unlike traditional AI models that typically classifies existing content, generative AI is much more creative and versatile.?

When applied to healthcare and biopharma, generative AI has the potential to impact how companies approach traditional processes.?

Use case: BioGPT for medical affairs?

No alt text provided for this image
Microsoft: BioGPT on PubMed v Human Performance



A particular form of AI that has huge potential for medical affairs is BioGPT.?

Microsoft recently announced the launch of #biogpt , which is a domain-specific generative model, trained on over over 15 million pieces abstracts and articles from #pubmed . This means that BioGPT is able to analyse biomedical research, answer questions, extract relevant data and generate text based on biomedical literature.?

According to Microsoft Research BioGPT can perform at the level of human experts and help researchers gain new insights in scientific discovery, including drug development.?

https://twitter.com/MSFTResearch/status/1618647707135918088 ?

BioGPT has many potential applications in biopharma. From creating new compelling medical content using existing resources, and tailoring this to different audiences, to answering queries on existing treatments or clinical studies, the use of generative AI will enable medical affairs to streamline workflows and better engage with healthcare professionals.?

#deeplearning and #medicalresearch ?

We have so far seen how different branches of AI can accelerate clinical trials and diagnostics or assist with generating high-quality medical content, but what is next??

Traditional machine learning approaches rely on both AI and human expertise to effectively and precisely analyse data.?

Deep learning is a cutting-edge subset of machine learning, which unlike other types of machine learning, can make decisions with significantly less human intervention.?

Deep learning is composed of a neural network with multiple layered algorithms, designed to continually analyse data in a logical structure. This more advanced branch of AI has advantages that can be applied to medical imaging, precision medicine, electronic health records and drug discovery.?

No alt text provided for this image
Deep learning's role in drug development process


Use case: Deep learning, drug discovery and precision medicine?

New deep learning-based approaches within each stage of drug development are becoming more apparent, particularly with the prediction of drug-drug interactions, drug properties and side effects . Although the drug development process is long and complex, deep learning can help to generate a compound with desired properties, while saving resources and accelerating the process.?

#precisionmedicine medicine is another area where deep learning models can have substantial impact. From expediting the process of analysing genomic and clinical data , to detecting and recommending the best possible treatment options for different patients.?

Exscientia is using deep learning ex vivo drug screening with patient tissue to identify personalised cancer treatments. Their EXALT-1 study demonstrated improved outcomes for late stage haematological cancer patients , using their AI-guided precision medicine platform that can recommend therapy options for individual patients.?

Deep learning offers promising approaches to drug discovery, precision medicine and improving patient outcomes.?

Considerations?

As AI becomes increasingly prevalent in healthcare and biopharma, there are a number of ethical, clinical & social challenges to consider before implementing AI.?

#privacy ?

With access to and the use of patient data, the appropriate safeguards must be in place to ensure the protection of patient data. Privacy concerns to consider include:?

  • Data security: AI algorithms are reliant on large volumes of patient data, which is at risk of being compromised. Biopharma must prioritise secure storage and transmission of patient data to mitigate this risk.?
  • Patient consent: As patients give their consent for their data to be used in AI applications, biopharma companies must ensure this data is only used for the intended purpose and be transparent about its usage.?
  • Bias: AI algorithms must be trained on diverse and representative datasets to prevent bias and discrimination.?

No alt text provided for this image
NIST Artificial Intelligence Risk Management Framework


#dataquality ?

AI applications are only as good as the data they are trained on. Quality data is needed for accurate predictions and reliable conclusions, particularly in healthcare. Data quality issues for AI in healthcare and biopharma include:?

  • Data accuracy: Inaccurate or incorrect data can lead to flawed conclusions and predictions.?
  • Data completeness: Large data sets are important for training, validating and improving algorithms. Without sufficient data, this poses the risk of unreliable and inaccurate algorithms providing incorrect conclusions.?
  • Data interoperability: Healthcare data is known for its heterogeneity, with varying formats or standards, which can make it difficult to integrate data into a single AI model.?

No alt text provided for this image
IBM Watson - Data Quality

#Auditability ?

With the use of AI applications multiplying within healthcare and biopharma, auditable AI is now essential. Auditable AI ensures that it is feasible to explain how AI models reach decisions and the factors that contribute to making decisions. This is particularly important in healthcare for the following reasons:?

  • Transparency: Auditing AI allows for greater transparency, helping stakeholders to understand how the AI arrived at its conclusions and the data it used. Transparency is crucial for building trust in AI and to verify and validate insights produced.?
  • Accountability: Accountability ensures decisions generated via AI align with ethical and legal standards. If issues arise, an audit trail can help determine where and how the AI failed and who is responsible.?
  • Quality Assurance: To ensure the quality of the AI, regular audits identify areas where the AI is underperforming or making mistakes. This information can be used to improve the AI's algorithms for optimal functioning.?

No alt text provided for this image
Source: Ryan Carrier

#HumanAdoption ?

Although the widespread adoption of AI has the potential to revolutionise the healthcare industry, there are certain challenges to the human adoption of AI in biopharma:?

  • Resistance to change: Healthcare professionals may be hesitant to adopt new technologies, especially if they perceive AI as a threat to their jobs.?
  • Lack of understanding and trust: Understanding how AI truly works and its benefits for biopharma is still not widely known. How AI models reach decisions may also not be trusted.?
  • Ethical concerns: There are several ethical concerns associated with the use of AI in healthcare, from data privacy to bias and discrimination. Adherence to guidelines and access to training can reduce concerns, by showing that AI is used in an ethical and responsible manner.?

No alt text provided for this image
Source: McKinsey Adoption of AI by Business Function & Industry


Final thoughts?

AI is being used to tackle some of the most pressing challenges in healthcare and biopharma, while reducing costs, improving efficiency and above all — improving patient care. Like all new technologies entering healthcare and biopharma, the topic of AI application has stirred debate on whether it is really safe to be used, with concerns and scepticism particularly around privacy. AI needs to be guided and humans will always need to oversee its usage, but undeniably, AI will continue to play a critical role in redefining the biopharma and healthcare industries.?

ChatGPT may be generating a lot of current interest and is certainly a major step forward, however this should be viewed through the prism of the historical achievements and near-term potential of a broader set of AI tools and approaches in revolutionising healthcare and biopharma.

The very human fears and issues are proving somewhat of a barrier to the pace of change and we must focus on addressing these, and the operational considerations in full to ensure we all benefit from the impact of this wave of technology in full.

要查看或添加评论,请登录

社区洞察

其他会员也浏览了