The Application of Artificial Intelligence in Rare Diseases

Consider These Complications with Rare Diseases:

1. Diagnosis is difficult and time-consuming, which means patients often spend a lot of time travelling through the healthcare system seeking a diagnosis. Patients are likely to see multiple doctors and try different treatments as they attempt to diagnose the disease. Years can lapse as doctors try to find a solution. First, they need to make the right diagnosis, and then, find the right treatment. Around 95% of rare diseases do not have a single treatment.
   
   
2. Because of the small numbers of patients, it is difficult to arrange appropriate care for them. Specialized units sometimes exist but are often underfunded.
   
   
3. Even when a treatment is known, it’s very costly as the cost to bring the drug to market is still roughly $4-12 billion. As you know, a rare disease means fewer patients with which to recover the costs of development, yet a customer base is the only way companies can fund the drugs. In many cases, despite high expense, these drugs are very effective. A European study found the cost-effectiveness of treatment for Gaucher Disease patients works out over a lifetime of a patient to be roughly Euro 885,000 per quality adjusted life years. Not bad. And the treatment for Fabry Disease also showed similar cost-effectiveness ratios. Although the cost to payers is relatively high per person, it is a small cost for the impact on the person’s life and a small cost compared to the cost to the healthcare system if they are not diagnosed.

So, where does Artificial Intelligence (AI) fit in?

Here Are Some AI Applications As Applied To Rare Disease:

R&D is a natural fit for AI and it’s applied in numerous ways in medical and genetic research.

Types of applications include:

• 3D tissue modeling to assess toxicity can be done with computer simulation rather than clinical trials on animals or people. Not only does this prevent the need for people or animal subjects but it can also shorten the length of clinical trials.
  
  
• Precision medicine – With AI, we can analyze a patient’s cells and compare them to healthy individuals and current studies to better assess the possible disease. With the analysis of millions, or even trillions, of data points that Artificial Intelligence can coalesce in a short amount of time, patients can be accurately treated faster.
  
  
• Difficulty in finding patients for clinical trials:
  • Given many conditions are genetic, there are genetic markers which can be found in biomarker tests. As you know, establishing biomarkers is important in tracking the effectiveness of treatments.
    
    
• Difficulty in gaining market access and pricing constraints:
  • What’s crucial that is needed is threefold: health outcomes for patients, value for payers, and financial sustainability for manufacturers. These can be achieved using a similar model of pricing that AirBnB uses i.e. machine learning. It can be applied to the core factors (patient outcomes, payer value and manufacturer profit) to find a price that satisfies all components. This, of course, is not just for rare diseases – this can be applied in all drug pricing.
    
    
• Difficulty in finding patients and achieving early diagnosis:
  • An interesting application Eularis has found is creating facial recognition software to find patients. With several of the genetic conditions, distinct facial traits are able to be detected using this kind of software.

This is an example of how Eularis use AI and social media to hasten the diagnosis of children born with a rare genetic disorder. In many cases, children with rare conditions do not get diagnosed at all; however, if they do, it is often late and sometimes too late to impact the outcome of the child. Our goal for our clients with a rare disease drug is to find these children using AI, and help their parents recognize the signs of this rare disorder so they could get the medical attention and diagnosis they need. We noticed in many genetic rare diseases that there are some facial similarities due to the condition that can be noticed. So, by creating facial recognition software, Eularis can write algorithms for bots to trawl Facebook, Flickr and other online sites where photos are uploaded, and identify patients who match the facial characteristics of this rare disease. The AI algorithm tags the identified children and searches their parents’ posts for mention of symptoms connected to this disease.

When the algorithms find instances of the condition and keywords of parents discussing the children’s challenges, we link this to a cookie driven ad campaign (similar to the pet feeder example in my article about segmentation) that shows an ad for the targeted keywords when, and only when, the parents are discussing or searching around them specifically – not when they are doing their online grocery shop for example. It must be targeted and relevant at the time it is shown. This way, they are led to a website to educate the parents about the condition their child has, to help them diagnose the child themselves and bring that information to their physician. This way, they go to the doctor armed with important insights that help the doctor with a short-cut to a medical diagnosis and treatment plan.