Podcast Summary: Chasing Life
Episode: Can AI Hear Cancer in Your Voice?
Host: Dr. Sanjay Gupta (CNN Chief Medical Correspondent)
Guest: Dr. Yael Bunsusan, Director of University of South Florida’s Health Voice Center
Date: September 5, 2025
Overview
In this episode, Dr. Sanjay Gupta explores the remarkable potential of artificial intelligence (AI) to extract hidden health information from the human voice. Joined by laryngologist and voice surgeon Dr. Yael Bunsusan, the discussion dives into how vocal biomarkers—analyzed by AI—could someday help diagnose diseases ranging from cancer to neurological and metabolic conditions, all from the sound of your voice. The episode also unpacks the science, technology, ethics, and future implications of AI-driven voice diagnostics.
Key Discussion Points & Insights
The Voice as a Window Into Health
- Human Voices Are More Than Words
- Dr. Gupta highlights how voices not only express thoughts and emotions but can also reveal health status. (00:34)
- “They can signal how you’re feeling physically, mentally. They can conjure up memories…the voice alone can reveal secrets.” (Dr. Sanjay Gupta, 00:34)
- Personal Anecdote from Dr. Bunsusan
- Dr. Bunsusan shares how her own vocal injury as a teenager, and subsequent recovery, led her from singing to medical specialization in voice—the path that underpins her current work. (03:18)
- “…when you lose your voice or … the ability to communicate with people… that was really powerful to see what it can do to help rehabilitate the voice.” (Dr. Yael Bunsusan, 03:18)
Who Sees a Voice Specialist?
- Wide Range of “Voice Professionals”
- Not just singers: TV hosts, teachers, anchors—anyone relying extensively on their voice.
- Patients Include:
- People with throat cancers or benign lesions
- Those with neurological disease affecting speech (e.g., Parkinson’s)
- Cases of voice paralysis or postoperative complications (05:09)
What Can Voice Reveal About Disease?
-
Three Categories of Disease Detection (Dr. Bunsusan, 07:38):
- Direct Voice Box Disorders: E.g., vocal cord cancer, paralysis
- Neurological Diseases: Voice or speech changes before physical symptoms (e.g., Parkinson’s, Alzheimer’s, stroke)
- Internal Medicine/Chronic Diseases: AI can pick up subtle vocal cues tied to diabetes, hypertension, and even atrial fibrillation—signals often imperceptible to the human ear but identifiable to trained algorithms.
-
Fascinating Anecdote:
- Dr. Bunsusan could tell a friend was pregnant just by the “stuffy” sound of her voice—an effect of hormonal changes on sinus tissues. (05:57)
- “I could tell just from the sound, because when we’re pregnant, there’s more estrogen…we sound kind of congested…” (Dr. Yael Bunsusan, 05:59)
AI's Growing Role in Voice Analysis
- Use in Remote Medicine & Telehealth
- In telehealth contexts, AI-enabled systems could issue risk reports based on routine patient conversations. (10:17)
- “…if we can converse and then at the end…you get a report saying your patient has a high risk of Parkinson’s, or sounds really depressed, that’s really helpful.” (Dr. Yael Bunsusan, 10:17)
- Ambient Scribes: Transforming Medical Notes
- New “ambient scribe” AI listens and drafts medical notes during the doctor-patient encounter—this could also integrate vocal biomarker analysis for unbiased health assessments. (10:17)
- “…if you combine ambient scribes with vocal biomarker technology…at the end…they get a report: this is the clinical history, but also from the voice, this is what I hear…” (Dr. Yael Bunsusan, 10:17)
Distinguishing Voice from Speech
- Speech vs. Voice
- Voice is the sound from the vocal cords; speech incorporates articulation and cognitive processing.
- Both speed and intonation patterns can serve as disease markers (e.g., monotone in Parkinson’s, slowed speech in cognitive decline). (12:48)
- “Birds can voice, but they cannot speak…When we talk about how voice can talk about our health, we also include speech.” (Dr. Yael Bunsusan, 12:48)
Challenges and Cautions
-
AI Accuracy & Trust
- AI systems must be rigorously trained on vast, diverse data to account for variability (health status, languages, cultures, etc.).
- “…if you’re a doctor adopting technology, you want to be sure that it’s accurate. For it to be accurate, it has to be trained on a lot of different people.” (Dr. Yael Bunsusan, 20:00)
-
Risks of Misinterpretation
- Illness (e.g., laryngitis) might be mistakenly classified as stress unless training datasets are comprehensive.
-
Cultural and Linguistic Differences
- Early AI models struggled across languages, but newer large datasets and improved algorithms mean these barriers are less pronounced. (21:17)
-
Aging and the Voice
- Vocal cords and the sound of the voice change with age—pitches generally drop or rise and voices become breathier or weaker, sometimes blurring gender differences in the elderly. (22:06, 23:10)
- “So our vocal cords, as we age…there's some muscle atrophy…we’re going to have a breathier voice…usually a weaker voice.” (Dr. Yael Bunsusan, 23:10)
Notable Cases & Social Impact
-
Famous voices and conditions: Discussion of public figures like RFK Jr. and President Biden to illustrate vocal disorders.
- “He sounds like he has spasms in his voice…which we can associate with diseases called spasmodic dysphonia…It’s very debilitating…” (Dr. Yael Bunsusan, 24:08)
- Also highlights how public perception of age and weakness is often tied to voice.
-
Loss of voice as loss of identity:
- “I could tell you that in a day of clinic, there’s at least two or three patients that cry in my office because…it's our identity.” (Dr. Yael Bunsusan, 25:37)
The Future: Opportunities and Ethics
-
Next 5-10 Years
- AI-powered voice diagnostics could become everyday tools for telehealth, smart speakers, and home monitoring, helping especially those in remote or under-resourced settings. (26:00)
- “…maybe a little robot at your house that you can talk to. The technology is going to listen to us…help vocal biomarkers take a bigger place in the market.” (Dr. Yael Bunsusan, 26:00)
-
Ethical Considerations
- Dangers of voice cloning and privacy abuses.
- Teaming up with bioethicists to protect patients, ensuring innovation serves people positively. (26:00)
- “So we can do great things with technology and we can do bad things with technology. Our job is to make sure we fuel innovation for the good…but also protect patients.” (Dr. Yael Bunsusan, 26:00)
-
Early Detection and Medical Advocacy
- Voice AI won’t replace specialists, but could reduce delays to specialist care, leading to earlier detection of cancers and other diseases.
- “If we can get these technologies to people earlier…they wouldn’t have that much of a bad cancer. It would have been curable.” (Dr. Yael Bunsusan, 27:46)
-
Practical Advice for Listeners
- Persistent voice changes (3-4 weeks or more) should prompt medical attention—don’t ignore changes.
Notable Quotes & Memorable Moments
- “You sound like something’s wrong with your voice…I went to see a laryngologist…had vocal cord nodules…voice rehabilitation expert…and I run a voice center at USF.” (Dr. Yael Bunsusan recounting her own journey, 03:18)
- “I could tell just from the sound, because when we’re pregnant…we sound kind of congested.” (Dr. Yael Bunsusan, on detecting pregnancy via voice, 05:59)
- “Birds can voice, but they cannot speak.” (Dr. Yael Bunsusan, on difference between voice and speech, 12:48)
- “Our voice really, it’s our identity.” (Dr. Yael Bunsusan, 25:37)
- “You can also clone a voice. You can harm someone…So we can do great things with technology and we can do bad things with technology.” (Dr. Yael Bunsusan, 26:00)
- “Three weeks of voice change or more…you should seek medical attention.” (Dr. Yael Bunsusan, 28:56)
Timestamps for Important Segments
- 00:34: Dr. Gupta introduces how voices can reveal secrets and signal health status
- 03:18: Dr. Bunsusan shares personal story; how a lost voice led to her medical career
- 05:57: Anecdote of detecting pregnancy via vocal changes
- 07:38: Three main categories of disease detectable in voice
- 10:17: AI and telehealth—using voice as a clinical biomarker
- 12:48: Differentiating “voice” from “speech” in diagnostics
- 20:00: Challenges and ethical cautions with AI voice analysis
- 22:06 – 23:10: How voices change as people age; what “old” voices sound like
- 24:08: Discussion of spasmodic dysphonia (RFK Jr.) and public perception
- 25:37: Emotional impact and identity issues with lost/altered voices
- 26:00 – 27:46: The future—AI voice technology, ethics, and disparities in care
- 28:56: Medical advice—seek attention for prolonged voice changes
Summary Takeaways
- AI voice analysis is on the verge of transforming diagnostics, particularly for remote or underserved communities.
- Voice is an untapped biomarker, signaling a spectrum of health issues—from cancer to Alzheimer’s—with the promise of earlier intervention.
- The intersection of technology, ethics, and medicine is critical: patient privacy, cultural variation, and algorithmic training are all major considerations.
- If your voice changes and doesn't return to normal after 3-4 weeks, seek medical attention—voice health is a window to overall well-being.
“Our voice really, it’s our identity…we can do great things with technology and we can do bad things with technology…our job is to make sure we fuel innovation for the good of people and our patients, but also protect patients from the bad.”
— Dr. Yael Bunsusan (25:37, 26:00)
