AI Diagnostic Tool Accurately Detects Valvular Disorders Often Missed by Doctors

Doctors generally use stethoscopes to listen for the characteristic lub-dub sounds made by heart valves opening and closing. They also listen for less prominent sounds that indicate problems with these valves. But the human body is full of other noises, like the flow of blood, stomach rumblings, and breath sounds, which can easily mask the signs of valvular heart disease (VHD). Research reveals that only 44% of VHD cases are caught through regular stethoscope check-ups. This leads to delayed diagnoses, worsening health conditions for patients, and huge costs for the healthcare system. To improve on this, researchers have now developed a new diagnostic tool that uses a short burst of audio data to accurately identify VHD.

Researchers at Stevens Institute of Technology (Hoboken, NJ, USA) used a contact microphone to take 10-second sound vibrations directly from a patient's chest. This data was then analyzed by an AI model adapted from algorithms normally used in speech processing to separate overlapping voices. In this case, the algorithm works to isolate the specific sounds associated with different types of heart valve diseases. The AI system can quickly identify up to five different types of valvular issues in a single patient, even if more than one condition is present.

The AI tool can detect VHD with 93% sensitivity and 98% specificity, substantially reducing undiagnosed cases and limiting false positives. The results are given in a simple 5-digit code made of ones and zeros, indicating the presence or absence of specific VHD. What sets this diagnostic tool apart from previously used neural networks to detect VHD is the use of accelerometers instead of more complex and cumbersome machines. Not only is this technique more accurate, but it also has the potential for further refinement. The team aims to extend their approach to identify other cardiovascular conditions and hopes to integrate this technology into medical practices nationwide, making it easier to diagnose cardiac disorders.

“Most cases of VHD are missed because of human error — so we brought in AI to help the human,” explained Negar Ebadi, the principal investigator of the project. “Our current goal is to collect more data so we can begin to classify diseases by severity — so instead of showing that you have a particular valvular disorder, we could give a grade out of 10 describing how far the disease has progressed.”

Related Links:
Stevens Institute of Technology