TRODENT FEATURE STORY
DIAGNOSIS CERTAIN
Ostrow faculty members develop three-app system, fueled by a proprietary artificial intelligence algorithm, to assist clinicians with proper orofacial diagnoses.
BY JAMIE WETHERBE MA ’04
PROVIDING A PATIENT WITH THE PROPER DIAGNOSIS can be challenging, especially when a condition is solely symptom-based. To zero-in on the correct diagnosis, clinicians must conduct the right combination of assessment questionnaires, exams and tests.
As many as 14 percent of new patient encounters result in misdiagnosis, according to a study that appeared in the American Journal of Medicine. This statistic isn’t necessarily a reflection of clinician experience, but of the quality of the current diagnostic tools.
As an orofacial pain specialist, Ostrow faculty member Glenn Clark felt frustrated by the way current electronic medical record systems can make it difficult to document, diagnose and provide good patient care.
“I’d spend more time clicking through screens and hunting for the right boxes to check than actually listening to my patients,” says Clark, Ostrow’s associate dean of distance learning and telehealth. “Meanwhile, I was seeing too many patients who had been misdiagnosed, sometimes for years, often with worsening pain. The problem wasn’t that they were being treated by bad healthcare providers — it was that our current systems don’t support good diagnostic thinking.”
Many providers — from dentistry to neurology to primary care — need to synthesize complex information and consider multiple possibilities.
“We rely so much on the patient’s story and history,” says Anette Vistoso MS ’20, assistant professor of clinical dentistry, director of the Oral Care Precancer and Pain Clinic and director of distance learning and telehealth. “We needed a better way to gather the right information, so we didn’t miss important pieces for each patient’s diagnosis.”
Vistoso and Clark, along with a research team at Ostrow, developed a new AI-enabled digital assistant for clinicians, with successful outcomes recently featured in the Journal of the American Dental Association.
The Smart Note system is designed to seamlessly cover a patient’s encounter — from medical history collection to physical examinations and onward — to help clinicians streamline the clinical workflow, enhance diagnostic accuracy and improve patient monitoring.
In a study diagnosing 50 test cases across 10 different clinical diagnoses, this integrated system, which consists of three web applications, was 93 percent accurate, outperforming five other AI algorithms operating on completed clinical records.
It was particularly accurate as a tool to diagnose conditions with distinctive symptom patterns, like trigeminal neuralgia (more than 97 percent accuracy), and burning mouth syndrome (more than 98 percent accuracy).
“The AI Suggests, the Doctor Decides”
Providers in certain fields, including orofacial pain, often can’t rely on biomarkers, like lab tests or image scans, to find definitive answers. Success typically hinges on collecting comprehensive patient histories and performing targeted physical exams.
“I noticed that, even after decades of practice, I would sometimes forget to ask a crucial question that could change the entire diagnostic picture,” Clark says. “We strongly believe that if we had a better, more standardized diagnostic navigation system, patients would get more accurate diagnoses faster. Instead of bouncing between doctors for months or years, they get the right answer the first time.”
Accurate diagnosis through documentation is critical. But with increasing clinical demands and more than 100 different conditions to differentiate, even experienced orofacial specialists can miss critical details.
“We thought, ‘What if we could have an AI assistant that helps ensure we never overlook the important questions?’” Clark shares.
Using data provided by patients and clinicians, Smart Note uses a patented algorithm to point to the most useful next steps, allowing the provider to automatically navigate to the appropriate place in clinical records for documentation.
After the clinician confirms a diagnosis, the resulting record allows it to be added automatically to the AI database, so the system can increase in scope and accuracy as more clinicians and patients use it.
“Our system grows from real clinical experience,” Vistoso says. “Each suggestion comes from analyzed data based on a real patient population, representing a range of demographics.”
This level of curation is key to AI’s accuracy and keeps it from generating random diagnoses — also known as “hallucinating.”
“We also keep the human doctor in complete control,” Clark adds. “The AI suggests; the doctor decides.”
The Smart Note system evolved from Vistoso’s thesis project, which she started alongside Clark in 2016.
“I was trying to develop this new note-taking system to improve our clinical templates,” Vistoso explains. “We recognized the fundamental need to ask the right questions to identify the correct disease, and the need to know the variables and information for each condition. That’s how we started developing our own system.”
The Human Connection’s Essential
A patient’s encounter with this three-prong system starts with a comprehensive health form ahead of their appointment. Augmented by AI, the smart questionnaire is designed to feel more like a conversation, asking follow-up questions based on the patient’s answers.
“Often, patients come with a referral, and they don’t know exactly what for,” Vistoso says. “This gives them an avenue to tell us about their symptoms, their pain and chief complaints — instead of relying on the referring clinician, whose concerns might not match the patient’s.”
With a more complete picture, providers can ask targeted questions and perform a more thorough exam.
During the appointment, clinicians enter information into the second app — the Smart Note system, which immediately suggests to the clinician which tests or potential diagnoses to explore.
“For instance, if a patient’s symptoms suggest trigeminal neuralgia, it might prompt me to test for trigger zones I might otherwise forget to check,” Clark says.
The third and final phase focuses on the follow-up time period, typically an undocumented time between appointments. The MyDocNote app aims to fill the gap between clinical visits by prompting patients with regular reminders to answer questions about their treatment.
This provides clinicians with real-time insights into treatment efficacy, potential side effects and the need for follow-up visits. The patient-reported data funnels into Smart Note, creating a comprehensive view of the patient’s condition and treatment response over time from their point of view.
“The feedback from patients has been really positive; they feel they have a role in their own treatment,” Vistoso says. “It gives them a voice in their care and helps ensure patient concerns are heard and addressed promptly by the clinician.”
Specialists currently testing the technology have similar sentiments.
“They’re able to spend more time with patients and feel more confident in their diagnoses,” Vistoso explains. “It makes it very easy for them to communicate with clinicians from different specialties and the referring doctor.”
As for future use, Clark and Vistoso plan to add voice capabilities so doctors can simply speak their observations. They also aim to expand into other medical specialties that rely heavily on clinical reasoning for diagnosis — for instance, headache disorders.
While systems like Smart Note can be powerful tools in treatment planning and patient monitoring, Vistoso and Clark agree that AI could never replace doctors.
“It could never perform the physical examination or diagnostic tests,” Vistoso says. “Beyond that, a machine can’t read the patient, or empathize with their pain or feelings — that’s something only a human can do.”
Adds Clark, “AI will never replace the human connection that’s essential to healing.”
EASY AS 1, 2, 3
The Smart Note System consists of three components to assist healthcare providers with the ability to synthesize complex information and consider multiple possibilities when documenting, diagnosing and delivering excellent patient care.
1. Before Appointment
Before a patient’s first appointment, they fill out both a medical history questionnaire as well as a smart questionnaire. The interface is meant to feel like a conversation, with follow-up questions posed based on a patient’s answers.
2. During Appointment
During the appointment, clinicians enter data into a second app, called the Smart Note System, which synthesizes the information and suggests to the clinician which tests or potential diagnoses to explore.
3. After Appointment
The third component — the MyDocNote app — picks up after the visit, prompting regular follow-up questions to patients so that clinicians can better understand and ultimately predict clinical outcomes.
Byte-Sized Oral Health Education
Patient care is not the only area in which artificial intelligence is assisting Ostrow’s operation. In an attempt to better facilitate learning for younger generations, a series of podcasts have been created (consisting of full episodes, which are a deep dive into complex topics that can take up to 30 minutes, and briefs, quick two-minute summaries) using artificial intelligence and course materials. The idea is to meet Gen Z learners where they are. As products of the 21st century, this generation, born between 1997 and 2012, tends to expect flexible, on-demand, mobile-first content — so why shouldn’t their course materials be the same? To create these podcasts, instructors simply upload course materials into AI technology, which then creates conversational, engaging and easily accessible audio content that, once reviewed by an expert, can be edited into episodes and stored internally in one place for easy access. The first episode covered Caries Management by Risk Assessment.
You can listen to some clips here:
►Treatment Planning Principles