Chapter 2
AI to Enable Early Diagnosis and Support Clinical Decisions
A recent report found that medical errors were the third leading cause of death in the US [2]. Most of the errors are diagnostic, including misdiagnosis and diagnoses never given to patients. Most people in the United States experience at least one such misdiagnosis in their lifetime, and 10% of cases result in death [2] [3].
With AI, information processing and decision making become more efficient and less error prone. The examples below illustrate how AI-based devices can help healthcare providers deliver improved diagnoses directly from medical images, physiological signals, or patient health records.
An AI-based real-time map of the heart and its electrical activity helps doctors pinpoint surgical interventions for atrial fibrillation. (Image credit: Corify Care)
Challenge
There are almost one in three adults over the age of 65 who fall each year, making falling the main cause of fatal and nonfatal injuries in this age group.
Solution
Kinesis Health Technologies developed a device called QTUG™ (Quantitative Timed Up and Go), an objective, quantitative method of screening for fall risk, frailty, and mobility impairment using wireless inertial sensors placed on a patient’s leg. The final product uses AI-based models developed with MATLAB to compute a fall risk estimate (FRE) and a frailty index (FI).
- In a QTUG test, a patient is fitted with two wireless inertial sensors, one on each leg below the knee. Each sensor includes an accelerometer and a gyroscope.
- To remove high-frequency noise in the data collected from these sensors, they used digital filters they designed using the Filter Designer in Signal Processing Toolbox™.
- The team used Statistics and Machine Learning Toolbox™ to select the subset of features with the highest predictive value and validate a regularized discriminant classifier model implemented in MATLAB.
- The team trained their models on clinical trial data collected on thousands of patients and assessed the results produced by the combined classifier.
- To update the classifier coefficients based on a new reference data set, the engineers export them from MATLAB to a resource file that is then incorporated into their build.
Patient quantitative metrics. (Image credit: Kinesis Health Technologies)
Results
To date, clinicians in eight countries have used QTUG to evaluate more than 20,000 patients. The team continues to improve the reference data set as new results come in. The team estimates that its development time was three times faster than developing in Java®, reducing the time to market and registration as a Class I device with the US FDA, Health Canada, and the European Medicines Agency (EMA).
Challenge
A common arrhythmia in the heart is atrial fibrillation (AFib). When arrhythmia do not resolve with medication or shock therapies, patients may require invasive surgical ablation to disrupt problematic electrical signals and bring the heartbeat back to normal. The success rate of this procedure is only 50%, as targeting the right heart tissue is a challenge for the doctors.
Solution
Corify Care developed a device called Acorys® to provide a near real-time map of the heart and its electrical activity based on noninvasive signal measurements from the patient’s chest and back. Knowing the torso’s geometry and which electrical signals are coming from which part of the torso, one can work backward to reconstruct the heart.
The Corify team used MATLAB for machine learning and signal processing to filter noise, producing cleaner data that went through further signal processing to accurately reconstruct activity in the heart.
They trained their reconstruction algorithms with data from both patients and mathematical models, allowing Acorys to take electrical signals from any patient and convert that into a map of the heart.
Early prototype that reads signals from a person’s torso and reconstructs them into a map of the heart. (Image credit: Corify Care)
Results
The team is working toward obtaining a Conformité Européenne (CE) marking from the European Medicines Agency (EMA) and approval from the US Food and Drug Administration (FDA). By providing a detailed, noninvasive look at the heart and its electrical activity, Acorys has the potential to prevent unnecessary ablation procedures, saving patients and the healthcare system time and money.
Challenge
Cataract patients have blurred vision due to opacity that forms within the lens of the eye. Millions of people must resort to cataract surgery, where the natural lens is removed and replaced with an artificial intraocular lens (IOL). It is difficult for the ophthalmologist to accurately predict the power of the artificial lens needed for an optimal postoperative result.
Solution
An ophthalmologist named Dr. Warren Hill worked with the MathWorks team to create a novel radial basis function (RBF) to predict the optical power calculated for the intraocular lens.
- The RBF was developed by using MATLAB to train an AI-based model with detailed measurements of thousands of patients’ eyes before surgery using a Lenstar biometer.
- The training data also included observed postoperative outcomes.
- The team exported the model to Simulink, a graphical environment for designing, simulating, and testing systems, and then generated code from the model and deployed it to the Lenstar device.
Radial basis function (RBF) IOL calculator developed in MATLAB.
The calculator is known as the Hill-RBF calculator. An online version of this calculator was also released simultaneously, so any ophthalmologist worldwide would have access.
Results
The calculator, launched in 2016, was quickly adopted by the worldwide ophthalmic community. Using updated and refined versions of the calculator that incorporate a much larger data set, surgeons are now seeing outcomes at a 90% ±0.50 D accuracy, compared with the 78% success rate of the more commonly used traditional and older methods. To put that in perspective, with approximately 28 million surgeries conducted worldwide annually, a 12% improvement in outcomes would result in 3.4 million additional surgical successes.
References
[2] Makary, Martin A, and Michael Daniel. “Medical Error—the Third Leading Cause of Death in the US.” BMJ, May 3, 2016. https://doi.org/10.1136/bmj.i2139.
[3] Balogh, Erin, Bryan T. Miller, and John Ball. Improving Diagnosis in Health Care. Washington, DC: The National Academies Press, 2015.
