Difference between revisions of "AUDIOMICS"
(Audiomics aims to identify vocal biomarkers related to varying disease states) |
(Audiomics mines voice data for acoustic biomarkers to detect a broad range of diseases) |
||
| Line 5: | Line 5: | ||
Audiomics is an emerging subfield of [[bioinformatics]] that aims to identify vocal biomarkers related to varying pathological states. Audiomics combines vocal analysis with informatics approaches to collect and analyze biological data.1,2 | Audiomics is an emerging subfield of [[bioinformatics]] that aims to identify vocal biomarkers related to varying pathological states. Audiomics combines vocal analysis with informatics approaches to collect and analyze biological data.1,2 | ||
| − | Advances in the processing methodology and technologies used in data science have changed the landscape of biomedical research through the expansion of “omics” fields, such as genomics, proteomics, and transcriptomics. Omics | + | Advances in the processing methodology and technologies used in data science have changed the landscape of biomedical research through the expansion of “omics” fields, such as genomics, proteomics, and transcriptomics. Omics use computational methodology to study the large amounts of data that represent the structure of a biological system.1,3-5 Arguably, the omics wave began with achievements in mapping the human genome.3,5 Technology to obtain the massive number of micro-measurements required to sequence the human genome led to the ability to explore other biological systems that were not previously possible.6 Now, decades since the human genome project began, the omics approach to analyze large amounts of biological data has branched into audio. Audiomics mines voice data for acoustic biomarkers to detect a broad range of diseases, including a wide range of otolaryngeal pathologies, neurological and psychological disorders, cardiovascular and respiratory disorders, and diabetes.1,7 |
| Line 12: | Line 12: | ||
| − | Early forms of technology using voice data were mainly seen in automated phone systems | + | Early forms of technology using voice data were mainly seen in automated phone systems. These phone systems processed predetermined and prompted answers through an algorithm to guide the user through instructions and commands without the assistance of a live agent. The recorded responses are filtered through an algorithm, dissected into commands and sent back to the user as output.2 Current forms of voice-data technology, takes the form of voice-based assistants. Voice-based assistants work through smartphones or home devices such as Amazon’s Alexa, Apple’s Siri, Google Assistant and Microsoft’s Cortana. These devices all have capacity to answer basic questions, control media and other automated home devices and track daily tasks.2,8 And, widespread acceptance of voice-based technology continues to grow. In 2019, almost one-third of smart phone users worldwide used voice activated technology regularly. Widespread use has prompted innovative techniques for audio analysis, including machine learning and natural language processing methods and made expansion into other sectors inevitable.8 |
=='''Voice as insight to disease state'''== | =='''Voice as insight to disease state'''== | ||
| − | Vocal sounds are typically produced when the lungs send air past vocal cords. Sound is produced that is shaped by the tongue, lips and nasal cavities. The nervous system, along with the brain coordinates all of these processes to form speech. Any part of the speech process that is affected by disease may leave clues for analysis.9 For example, with heart failure, fluid accumulation in the lungs affects the needed respiratory output required for speech. This results in less stable | + | Vocal sounds are typically produced when the lungs send air past vocal cords. Sound is produced that is shaped by the tongue, lips and nasal cavities. The nervous system, along with the brain coordinates all of these processes to form speech. Any part of the speech process that is affected by disease may leave clues for analysis.9 For example, with heart failure, fluid accumulation in the lungs affects the needed respiratory output required for speech. This results in less stable vocalization with shorter phrasing and speech rates.7 |
Early work with biomarkers mainly focused on speech patterns with Parkinson’s disease.9 The rates of voice disorders in Parkinson’s patients can be as high as 89%.8 Parkinson’s patients often have weak voices caused by motor symptoms including tremors and muscle stiffness that also affect the muscles involved in speech. Parkinson’s symptoms can be inconsistent and difficult to assess and historically, required an exam by a movement disorder specialist to diagnose.9 Similarly, the Alzheimer’s disease process produces changes in the voice that can be detected in early stages of cognitive impairment. Alzheimer’s patients have distinct vocal features characterized by frequent use of filler sounds, as well as variations in pitch and rhythm that reflect changes in the patient’s coherence and responsiveness.8 In audiomics, the objectively measured biomarker will provide a signature feature that is associated with a clinical outcome. Computational techniques facilitate measurement and analysis to quickly process large amounts of data and detect abnormalities to distinguish patients with the outcome of interest. Eventual outcomes could be related to a diagnosis, severity grade or disease stage.9 | Early work with biomarkers mainly focused on speech patterns with Parkinson’s disease.9 The rates of voice disorders in Parkinson’s patients can be as high as 89%.8 Parkinson’s patients often have weak voices caused by motor symptoms including tremors and muscle stiffness that also affect the muscles involved in speech. Parkinson’s symptoms can be inconsistent and difficult to assess and historically, required an exam by a movement disorder specialist to diagnose.9 Similarly, the Alzheimer’s disease process produces changes in the voice that can be detected in early stages of cognitive impairment. Alzheimer’s patients have distinct vocal features characterized by frequent use of filler sounds, as well as variations in pitch and rhythm that reflect changes in the patient’s coherence and responsiveness.8 In audiomics, the objectively measured biomarker will provide a signature feature that is associated with a clinical outcome. Computational techniques facilitate measurement and analysis to quickly process large amounts of data and detect abnormalities to distinguish patients with the outcome of interest. Eventual outcomes could be related to a diagnosis, severity grade or disease stage.9 | ||
Latest revision as of 06:24, 11 May 2024
Contents
A new omics
Audiomics is an emerging subfield of bioinformatics that aims to identify vocal biomarkers related to varying pathological states. Audiomics combines vocal analysis with informatics approaches to collect and analyze biological data.1,2
Advances in the processing methodology and technologies used in data science have changed the landscape of biomedical research through the expansion of “omics” fields, such as genomics, proteomics, and transcriptomics. Omics use computational methodology to study the large amounts of data that represent the structure of a biological system.1,3-5 Arguably, the omics wave began with achievements in mapping the human genome.3,5 Technology to obtain the massive number of micro-measurements required to sequence the human genome led to the ability to explore other biological systems that were not previously possible.6 Now, decades since the human genome project began, the omics approach to analyze large amounts of biological data has branched into audio. Audiomics mines voice data for acoustic biomarkers to detect a broad range of diseases, including a wide range of otolaryngeal pathologies, neurological and psychological disorders, cardiovascular and respiratory disorders, and diabetes.1,7
Early voice data technologies
Early forms of technology using voice data were mainly seen in automated phone systems. These phone systems processed predetermined and prompted answers through an algorithm to guide the user through instructions and commands without the assistance of a live agent. The recorded responses are filtered through an algorithm, dissected into commands and sent back to the user as output.2 Current forms of voice-data technology, takes the form of voice-based assistants. Voice-based assistants work through smartphones or home devices such as Amazon’s Alexa, Apple’s Siri, Google Assistant and Microsoft’s Cortana. These devices all have capacity to answer basic questions, control media and other automated home devices and track daily tasks.2,8 And, widespread acceptance of voice-based technology continues to grow. In 2019, almost one-third of smart phone users worldwide used voice activated technology regularly. Widespread use has prompted innovative techniques for audio analysis, including machine learning and natural language processing methods and made expansion into other sectors inevitable.8
Voice as insight to disease state
Vocal sounds are typically produced when the lungs send air past vocal cords. Sound is produced that is shaped by the tongue, lips and nasal cavities. The nervous system, along with the brain coordinates all of these processes to form speech. Any part of the speech process that is affected by disease may leave clues for analysis.9 For example, with heart failure, fluid accumulation in the lungs affects the needed respiratory output required for speech. This results in less stable vocalization with shorter phrasing and speech rates.7
Early work with biomarkers mainly focused on speech patterns with Parkinson’s disease.9 The rates of voice disorders in Parkinson’s patients can be as high as 89%.8 Parkinson’s patients often have weak voices caused by motor symptoms including tremors and muscle stiffness that also affect the muscles involved in speech. Parkinson’s symptoms can be inconsistent and difficult to assess and historically, required an exam by a movement disorder specialist to diagnose.9 Similarly, the Alzheimer’s disease process produces changes in the voice that can be detected in early stages of cognitive impairment. Alzheimer’s patients have distinct vocal features characterized by frequent use of filler sounds, as well as variations in pitch and rhythm that reflect changes in the patient’s coherence and responsiveness.8 In audiomics, the objectively measured biomarker will provide a signature feature that is associated with a clinical outcome. Computational techniques facilitate measurement and analysis to quickly process large amounts of data and detect abnormalities to distinguish patients with the outcome of interest. Eventual outcomes could be related to a diagnosis, severity grade or disease stage.9
The national and regional lockdowns forced by the Covid-19 pandemic put pressure on healthcare systems to identify solutions for efficiency and alleviating clinical staff workload. Digital technology and Artificial Intelligence (AI) based solutions enabled patient self-surveillance and remote monitoring of symptoms towards these efforts.8,10 Vocal biomarkers became a promising source of data available at a large scale to employ digital technology and Artificial Intelligence (AI) based solutions to improve the healthcare workload. Voice data is a data source that is rich, user friendly, cheap and non-invasive to collect. Since vocal biomarkers had already been identified as a diagnosis tool in other contexts, there was precedent to use vocal biomarkers for Covid-19 symptom assessment.9,11,12
The Covid-19 pandemic highlighted a need to quickly identify personalized care options. While asymptomatic patients, or patients with mild or moderate symptoms only required management through home-based care, many patients with worsening symptoms would benefit from quick and objective assessments to determine risk for hospitalization. Based on the hypothesis that symptomatic Covid-19 cases would present different audio features compared to symptomatic cases, several research teams took to leveraging previous work with biomarkers and developed screening tools for covid based on recordings of respiratory sounds (cough, breathing, voice).8,9,11-12
Limitations to clinical use
The Covid-19 pandemic spurred many of the recent efforts to analyze vocal data for disease biomarkers. Subsequent reports show promising results for accurate detection and classification. 8,9,11-12 However, many tools are still in the pilot or proof of concept stage, and in spite of great potential, there is still a long road to widespread clinical use.1,10 The FAIR principles, established in 2016, state that biological research data should be findable, accessible, interoperable and reusable. Although the FAIR principles have established important guidelines for data management, there are still privacy and consistency issues across omics fields.13 As one of the newest omics, audiomics standards for audio collection, storage and analysis are still in development. Under the Health Insurance Portability and Accountability Act (HIPPA), voice is considered an identifier and regulated by data sharing stipulations.1 Data sharing stipulations also limit the collaborative efforts that are key to scientific progress. To comply, research teams maintain small datasets only used within the team’s academic institution. These factors, combined, have all prevented the comparative studies required to qualify for federal approval. The United States Food and Drug Administration (FDA), as well as the European Medicines Agency have yet to approve any vocal biomarkers for clinical use.10
Bridge2AI-Voice
To propel United States national efforts to establish data standards, in 2022 the National Institutes of Health pledged $130 million over four years towards establishing data standards, best practices and ethical considerations for the data used in artificial intelligence.14 The program funds four data generation projects. These databases will facilitate discoveries using AI and machine learning to advance research in diabetes, clinical care, genotype/phenotype learning and vocal biomarkers. The Bridge2AI Voice project is a consortium of institutions that will examine five disease categories (vocal pathologies, neurological and neurodegenerative disorders, mood and psychiatric disorders, cardiovascular and respiratory disorders, and pediatric diseases) using vocal biomarkers associated with screening, diagnosis and treatment of disease.15 The consortium is composed of voice AI experts and researchers with experiences as data and acoustic engineers, bio-ethicists, speech pathologists, clinicians and educators from twelve institutions across the US and Canada.15 The standardization efforts of Bridge2AI Voice will not only improve the accuracy of models used in audiomics, these efforts will also strengthen the future potential to contribute meaningful answers to clinical questions.
REFERENCES
1. Bensoussan Y, Elemento O, Rameau A. Voice as an AI Biomarker of Health—Introducing Audiomics. JAMA Otolaryngol Head Neck Surg [Internet]. 2024 Apr 1 [cited 2024 Apr 29];150(4):283. Available from: https://jamanetwork.com/journals/jamaotolaryngology/fullarticle/2815136 2. Gouda P, Ganni E, Chung P, Randhawa VK, Marquis-Gravel G, Avram R, et al. Feasibility of Incorporating Voice Technology and Virtual Assistants in Cardiovascular Care and Clinical Trials. Curr Cardiovasc Risk Rep [Internet]. 2021 [cited 2024 Apr 29];15(8):13. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8214838/ 3. McColl ER, Asthana R, Paine MF, Piquette-Miller M. The Age of Omics-Driven Precision Medicine. Clinical Pharmacology & Therapeutics [Internet]. 2019 [cited 2024 May 3];106(3):477–81. Available from: https://onlinelibrary.wiley.com/doi/abs/10.1002/cpt.1532 4. Dai X, Shen L. Advances and Trends in Omics Technology Development. Front Med (Lausanne) [Internet]. 2022 Jul 1 [cited 2024 May 3];9:911861. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9289742/ 5. Evans GA. Designer science and the “omic” revolution. Nat Biotechnol [Internet]. 2000 Feb [cited 2024 May 3];18(2):127–127. Available from: https://www.nature.com/articles/nbt0200_127 6. Micheel CM, Nass SJ, Omenn GS, Trials C on the R of OBT for PPO in C, Services B on HC, Policy B on HS, et al. Omics-Based Clinical Discovery: Science, Technology, and Applications. In: Evolution of Translational Omics: Lessons Learned and the Path Forward [Internet]. National Academies Press (US); 2012 [cited 2024 May 3]. Available from: https://www.ncbi.nlm.nih.gov/books/NBK202165/ 7. Murton OM, Dec GW, Hillman RE, Majmudar MD, Steiner J, Guttag JV, et al. Acoustic Voice and Speech Biomarkers of Treatment Status during Hospitalization for Acute Decompensated Heart Failure. Applied Sciences [Internet]. 2023 Jan 31 [cited 2024 Apr 29];13(3):1827. Available from: https://www.mdpi.com/2076-3417/13/3/1827 8. Fagherazzi G, Fischer A, Ismael M, Despotovic V. Voice for Health: The Use of Vocal Biomarkers from Research to Clinical Practice. Digit Biomark [Internet]. 2021 Apr 16 [cited 2024 Apr 29];5(1):78–88. Available from: https://www.karger.com/Article/FullText/515346 9. Anthes E. Alexa, do I have COVID-19? Nature [Internet]. 2020 Sep 30 [cited 2024 Apr 29];586(7827):22–5. Available from: https://www.nature.com/articles/d41586-020-02732-4 10. Fagherazzi G, Zhang L, Elbéji A, Higa E, Despotovic V, Ollert M, et al. A voice-based biomarker for monitoring symptom resolution in adults with COVID-19: Findings from the prospective Predi-COVID cohort study. McGinnis RS, editor. PLOS Digit Health [Internet]. 2022 Oct 20 [cited 2024 Apr 29];1(10):e0000112. Available from: https://dx.plos.org/10.1371/journal.pdig.0000112 11. Silva G, Batista P, Rodrigues PM. COVID-19 activity screening by a smart-data-driven multi-band voice analysis. J Voice [Internet]. 2022 Nov 15 [cited 2024 Apr 29]; Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9663738/ 12. Laguarta J, Hueto F, Subirana B. COVID-19 Artificial Intelligence Diagnosis Using Only Cough Recordings. IEEE Open Journal of Engineering in Medicine and Biology [Internet]. 2020 [cited 2024 Apr 29];1:275–81. Available from: https://ieeexplore.ieee.org/document/9208795 13. Oestreich M, Chen D, Schultze JL, Fritz M, Becker M. Privacy considerations for sharing genomics data. EXCLI J [Internet]. 2021 Jul 16 [cited 2024 May 2];20:1243–60. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8326502/ 14. Britt R. National Institutes of Health (NIH). 2022 [cited 2024 May 2]. News Release - NIH launches Bridge2AI program to expand the use of artificial intelligence in biomedical and behavioral research. Available from: https://www.nih.gov/news-events/news-releases/nih-launches-bridge2ai-program-expand-use-artificial-intelligence-biomedical-behavioral-research 15. Bridge2AI - Voice [Internet]. [cited 2024 May 2]. Available from: https://www.b2ai-voice.org/
Submitted by (Ayo Babatunde)
