These appeared last week:
Top AI healthcare stocks - Part 1
By streamlining workflows and improving diagnostics, Pro Medicus is the Apple of AI healthcare.
By Graham Witcomb · 5 Jul 2024
Practically all medical fields will benefit from artificial intelligence (AI) but, for now, most applications are more science fiction than fact. The medical imaging industry, however, is already experiencing the advantages — and Pro Medicus is leading the charge.
Currently, most algorithms approved for use by the US Food & Drug Administration (FDA) are focused on medical imaging. Plenty more will come: traditional radiology relies heavily on pattern recognition, which is an area where AI performs exceptionally well.
Key Points
· AI to improve diagnosis
· Opening platform to AI ecosystem the best bet
· Stock priced for growth
AI has two main applications within medical imaging — prioritisation and diagnosis.
AI is being used to triage cases by reading through the endless stream of scans and putting those patients with the highest risk of pathology at the top of the pile for a radiologist to review.
One German university hospital found that using AI to prioritise cases reduced the average turnaround time for reporting critical findings from 80 minutes to 35-50 minutes. This may not sound like much, but there's a major shortage of radiologists in Australia, Europe, and the US, so small increases in efficiency mean radiologists can see more patients and reduce wait times.
Diagnostic accuracy
So far, Pro Medicus's bread and butter has been a software platform that processes medical images remotely and streams the necessary pixels to the radiologist's viewing station, allowing doctors to access high-resolution scans almost instantly. This approach improves efficiency and output, enabling faster diagnosis and more convenient viewing (see Analyst picks: Uncovered gems).
While superior processing speed is Pro Medicus's main competitive advantage — allowing it to charge a 50% premium to competitors — the company is investing in machine learning to enhance its software's capabilities, including the ability to spot anomalies and assess patient risk.
The company has taken a shotgun approach to AI. Its three-pronged strategy includes: (1) developing its own AI algorithms internally; (2) partnering with third-party providers and universities to co-develop them; and (3) providing an open imaging platform accessible to independent AI developers.
The first prong of that strategy is arguably the weakest. In 2021, Pro Medicus received FDA approval for an internally developed algorithm to detect and categorise breast density. Since then, however, the company has barely said a word about internally developed AI — there may be further advances, but we're more excited by a number of partnerships with external developers.
Earlier this year, Pro Medicus invested US$5m into Elucid, a US-based AI provider that specialises in a Fractional Flow Reserve (FFR) evaluation tool for cardiac CT scans, which measures the pressure gradients in a coronary artery. The procedure's use is limited by its cost, complexity, and invasiveness, so an AI-driven non-invasive estimation of FFR will be a major benefit to patients.
Elucid has raised around US$120m in total funding. Pro Medicus's stake is small, but it shows the company's seriousness about building relationships with AI providers and funding their growth. Pro Medicus intends to partner with Elucid and integrate its AI algorithms into Pro Medicus's imaging platform.
Apple-esque
While developing AI algorithms and partnering with start-ups are both opportunities worth exploring, our best hopes are pinned on the third prong in Pro Medicus's AI strategy — its imaging platform.
Most revenue currently stems from image processing, not charging for AI services, and Pro Medicus boasts impressive financials: gross margins over 99%, with a free cash flow margin and return on equity each brushing 50%. Over the past three years, revenue from imaging has grown by 30% a year. Anything that strengthens the imaging platform's competitive position is a good thing.
Rather than focus solely on building the best AI models, the company has opened its imaging platform to third parties by providing accessible APIs (application programming interfaces). These APIs allow other AI developers to easily integrate their AI models into Pro Medicus's software.
We think it's the right strategy. Pro Medicus's strength is in speed and convenience, not AI development and diagnosis; open access is a way to use external developers — who may be better at AI than Pro Medicus — to improve Pro Medicus's core product.
Two things are clear: Pro Medicus's platform shows the output of AI, which is ultimately what matters to the radiologist; and there are hundreds of companies working on AI algorithms to improve diagnosis across an endless array of diseases. Given the choice between using dozens of different applications or using a single platform that shows the output of dozens of different AI models, we suspect the latter will win out.
Just as Apple's success with the iPhone stemmed from providing a platform for apps, Pro Medicus's approach allows radiologists to access a wide range of AI models through a single, streamlined interface.
It's hard to fault Pro Medicus's strategy and, so far, the company has gone from strength to strength. With roughly $150m of revenue, Pro Medicus has a 7% market share in North America, double what it was three years ago.
Unfortunately, the market has already caught on to Pro Medicus's potential: the stock's market cap of $14bn is 177 times the $80m of net profit expected in 2024. Consensus estimates are for net profit to more than double by 2028, making that price-to-earnings ratio more digestible — but anything other than rapid growth would still clobber the share price.
We'd love to own Pro Medicus at the right price, but we'll probably need to wait for a disappointing result before we get an opportunity.
For companies like Pro Medicus, AI is a core component of their product, making it integral to the user experience; other companies benefit from AI's capabilities quietly in the background. In Part 2, we'll explain how one such company's AI investments will deliver improvements in efficiency, costs, and research.
Here is the link:
https://www.intelligentinvestor.com.au/investment-news/top-ai-healthcare-stocks-part-1/153685
Top AI healthcare stocks - Part 2
No health stock has more to gain from AI than CSL, with drug and vaccine development just the start.
By Graham Witcomb · 10 Jul 2024 · 5 min read
In Part 1, we looked at how AI is transforming the medical imaging industry through the innovations of Pro Medicus. Now, we turn our attention to CSL, the undisputed leader in blood products and a top-tier producer of vaccines.
AI's potential applications within CSL are extensive, offering significant improvements in efficiency, speed, and research.
Key Points
· Predictive drug design
· Faster vaccine development will improve effectiveness
· Manufacturing efficiencies
Traditionally, drug discovery has been a lengthy and expensive process.
The first phase of drug development involves identifying potential drug candidates through laboratory studies to evaluate the biological activity and efficacy of these contenders. A bottleneck in this research is the sheer quantity of proteins available for study: blood plasma contains more than 4,000 different proteins, sometimes in indescribably small quantities, each of which could have an important biological effect when concentrated and administered to patients for a specific illness.
AI can help to prioritise research. By analysing huge datasets, algorithms can predict how different molecules will interact with biological targets, identifying promising candidates more quickly than conventional methods based on trial and error.
AI can analyse research data more precisely than humans to find unusual patterns or similarities in drug performance. For CSL, this means a more efficient path to discovering new therapies, reducing time and cost.
R&D accelerator
Once a drug candidate shows promise in basic research studies, it progresses to clinical trials, which are conducted in three phases to test the drug's safety, dosage, and side effects. The clinical trial sequence is the most expensive part of research and development (R&D), often costing hundreds of millions and taking more than a decade.
CSL hit this reality check with its plasma-derived cholesterol drug CSL-112. Earlier this year, the company announced the results of an enormous 18,000-person Phase III clinical trial and found that CSL-112 was no better than placebo at reducing cardiovascular events following a heart attack. CSL spent $500m or so and a decade of research on this single project.
As mentioned earlier, AI is likely to help in the drug discovery process by proposing molecules worthy of investigation or spotting anomalies in data. Its most valuable contribution, however, may be eliminating low-potential drugs earlier in the clinical trial process. Only 1-in-10 drugs that move from basic research to clinical trials make it through all three phases to commercialisation. R&D is a lottery where most research spending is wasted.
The final Phase III trial typically accounts for around 60% of clinical trial costs. By more thoroughly analyisng the vast quantities of data produced by earlier trials — or trials of similar molecules — AI may be able to predict the likelihood of success with more accuracy, helping CSL to allocate resources more effectively.
When one of CSL's therapies is finally approved, the research doesn't end there — the company is constantly looking to repurpose drugs, aiming to find new uses beyond their original indications. Most rare diseases — CSL's specialty — don't have any approved treatments, so AI has the potential to help CSL expand the market for its existing products by finding new therapeutic uses for them. Given the cost of development, an expanded list of 'off label' uses could add meaningfully to CSL's return on investment.
Vaccine development
AI's fast analysis is especially helpful for CSL's vaccine division. CSL has a 26% share of the flu vaccine market, making it the world's second-largest flu vaccine maker.
Influenza viruses mutate frequently, leading to a phenomenon known as antigenic drift. Small changes in the virus's surface proteins can lead to new strains that existing vaccines may not protect against.
The issue is that the flu virus is evolving constantly, whereas vaccine development takes many months. As soon as a new flu vaccine hits the shelves, it is already out of date because the virus has shifted slightly since it was developed.
To address this, flu vaccines are reformulated each year based on predictions of which strains will be most prevalent. This prediction process involves global surveillance and data collection by health organisations. Despite these efforts, there is almost always some degree of mismatch between the vaccine strains and the circulating strains, resulting in reduced vaccine effectiveness for that season.
CSL can use AI to model viruses and predict immune responses, speeding up vaccine design. By accelerating the development process, vaccine strains will more closely match the virus they're targeting.
AI is likely to become a standard feature in flu vaccine development, so we doubt it will give CSL any competitive advantage or increased profitability, but you never know — small differences in software performance could lead to large discrepancies in real-world outcomes. In a field where effectiveness is paramount, doctors are likely to recommend the most effective vaccine, so the company that delivers that could dominate the market. And more effective vaccines overall could improve demand.
Manufacturing
Two final applications of AI that could supercharge CSL's operations extend beyond R&D into manufacturing and compliance.
The company intends to use AI to improve the efficiency of its supply chain and assist in meeting its regulatory, legal, and compliance requirements. Predictive maintenance algorithms keep production equipment running efficiently, reducing downtime and increasing output. Real-time monitoring through machine learning models can detect anomalies early, preventing potential quality issues.
AI can aid CSL in making more informed strategic decisions, too. Predictive analytics can forecast market trends, assess demand, and guide management in where it invests the company's cash pile.
Management expects net profit of US$2.9bn-3.0bn in 2024, up 13 -17%, putting the stock on a price-to-earnings ratio of 32 at the midpoint.
AI has the ability to enhance almost every aspect of CSL's operations — from drug discovery and expanded labelling to vaccine development, manufacturing and quality control.
CSL's size and 100-year operating history is a major competitive advantage as it has troves of research data on which is can train AI models more effectively. The company spends 9% of revenue on R&D and its US$1.2bn research budget is bigger than almost any competitor.
With significant financial resources, CSL can invest in cutting-edge AI technologies and infrastructure, hire top talent, and collaborate with leading AI research institutions. We can't think of a healthcare stock better positioned to benefit from developments in AI. HOLD.
Here is the link:
https://www.intelligentinvestor.com.au/recommendations/top-ai-healthcare-stocks-part-2/153690
It is interesting that the review did not extend to some of the smaller AI companies in the health sector of which there are many in various states of development.
There is also a useful review here:
AI is already being used in healthcare. But not all of it is ‘medical grade’
Dean, School of Computing Technologies, RMIT University, RMIT University
CEO, Australian e-Health Research Centre, CSIRO
Professor of Medical Informatics, Macquarie University
Artificial intelligence (AI) seems to be everywhere these days, and healthcare is no exception.
There are computer vision tools that can detect suspicious skin lesions as well as a specialist dermatologist can. Other tools can predict coronary artery disease from scans. There are also data-driven robots that guide minimally-invasive surgery.
To precisely diagnose diseases and guide treatment choices, AI is used to analyse patients’ genomic and molecular data. For instance, machine learning has been applied to detect Alzheimer’s disease and to help choose the best antidepressant medication for patients with major depression.
Deep learning methods have been used to model electronic health record data to predict health outcomes for patients and provide early estimates of treatment cost.
Much more here:
We just have to wait and see how this evolves – as we can be sure there will be a lot of investment in this area over the next few years!
Watch this space….
David.