Periodic Reporting for period 1 - PhonLab (A clinical phonetics laboratory service (PhonLab): diagnostics of speech impairment via the web)
Reporting period: 2017-06-01 to 2018-11-30
Summary of the context and overall objectives of the project
The idea.
The objective of the ERC-Proof of Concept was to establish a commercial online service for the assessment of speech disorders. The service, named ‘PhonLab’, is planned to offer objective parameters over the web about the severity and specific pattern of voice and articulation impairment, given a sample of the patient’s speech. This is analogous to information delivered from a central clinical laboratory to a medical doctor about, for instance, cell changes, given a sample of the patient's tissue. In our case, the information includes diagnostic parameters concerning factors such as the degree of a patient's speech fluency or intelligibility obtained from an automatic analysis of the patient's acoustic speech signal provided via a web service. The results of this analysis are then returned to the user in the form of easily interpretable standard scores and graphics which provide clinicians with objective and sensitive data concerning treatment indication, treatment outcome, and the course of a speech disorder.
By providing high-end clinical and technological expertise independently of regional and social boundaries, PhonLab will enhance diagnostic standards for a large number of children and adults with speech impairments, initially in the German speaking countries, but with the longer-term perspective of a European and a world-wide service.
Resources.
PhonLab draws on techniques for processing child and adult speech acoustically that have been developed within the applicant's ERC project 'Sound Change and the Acquisition of Speech', as well as on research into the nature and treatment of speech disorders conducted at the host institute of the advanced ERC grant (IPS). The speech processing software developed at the IPS encompasses automatic speech segmentation tools, algorithms for the computation of all kinds of features describing a speaker’s voice and articulation, and tools required for the online recording of speech signals and the management of large speech databases. The clinical expertise available at the host institute includes experience with a wide range of clinical conditions and knowledge about the diagnostic validity and usability of acoustic speech parameters. Furthermore, a pilot system for web-based auditory assessment of speaker intelligibility has already been hosted and entertained at the IPS. The project’s main objective was to exploit these achievements and bring PhonLab to the point of initiating a start-up process towards commercialisation.
Beneficiaries.
The system, once established, will have extensive benefits to adult patients with neurological disorders of speech and language (stroke, Parkinson’s disease, Multiple sclerosis, etc.), children with acquired or developmental speech-language disorders, and children or adults with lesions of the speech organs, i.e. mouth, throat, larynx, and respiratory organs. The prevalence of these disorders with their devastating consequences for communication and participation amounts to an estimated 4 million patients in Europe.
The clinical caregivers of these patients, i.e. speech-language therapists working in clinics and private practices, will benefit from the cost-effectiveness and free availability of a service providing them quantitative and objective parameters they can use to improve their treatment planning and disease monitoring. Ultimately, the health care system and its stakeholders benefit from the cost-efficient provision of equal access to cutting-edge speech processing technologies across regional and economic boundaries.
The use of advanced speech technology and digital telematics for the clinical assessment of communication disorders is in line with efforts in Europe and the rest of the world to establish health telematics as a measure to enhance the efficiency of healthcare provision.
Challenges.
The following major obstacles were to be overcome in the development of PhonLab as a web-based service:
(1) The system must comply with EU regulations for Medical Device Security, including, among other things, data protection and quality management requirements.
(2) It must be scalable regarding the volume of orders per day/week, the scale of diagnostic parameters offered (e.g. voice, resonance, articulation, fluency etc.), the range of diagnostic indications (e.g. neurologic, developmental, ENT), the clients’ business forms (e.g. private practice, clinical unit), the specific documentation requirements of health providers and insurance companies, and the language in which assessment materials are presented.
(3) It must generate valid, reliable and customized diagnostic parameters in a most cost-efficient way, i.e. under conditions of almost complete automation.
Technical solution.
Building on experiences with an existing pilot system based on auditory instead of acoustic analyses, PhonLab was designed as a modular “Software-as-a-Service” (SaaS) system with the following components:
- A device for the flexible generation of assessment materials (words, sentences) from a large German corpus annotated for lexical parameters (word length and phonological transcript, syllabicity, word-type, lexical frequency),
- a browser-independent speech recording device which is designed to bypass the automatic speech processing routines implemented in the Capture-Device tools of most browsers,
- a transport system guaranteeing a secure transfer of speech samples from the end-user to the PhonLab-server by automatic “TLS-encryption”,
- a scalable and secure user management component allowing for a specification and management of various user types (patients, therapists, listeners, speech analysis experts, laboratory assistants, administrators),
- a tool chain encompassing programs for the automatic segmentation of speech samples and the extraction and analysis of local and global features characterizing a patient’s speech impairment, according to clients’ requests.
Privacy and medical device requirements.
Speech data acquisition, transport and storage, and the management of user access were designed to conform to the General Data Protection Regulation of the EU. Contracts of PhonLab with their clients (i.e. clinics, therapists) and of clients with their patients will be drafted according to legal stipulations, following the advice provided by an external data protection officer.
The requirements for certification as a medical device of security category IIa have been considered and documented on each step of the development taken so far. An application for certification will be submitted as soon as the system is ready for spin-off.
Market and competitors.
PhonLab’s target market at entry level encompasses speech and language disorders in children and adults in the German speaking countries, with an overall volume of more than 30 m Euros to date spent for diagnostics in Germany alone. Target users are speech-language therapists (28,000 in Germany) who work in ca. 10,000 units (clinics, private practices) and treat more than 850,000 patients per year. Further market development will expand on other language communities in Europe and world-wide as well as on fields of application outside the health sector, such as pronunciation training and assessment for migrants working in foreign language environments.
If acoustic speech analysis methods play a role at all in clinical diagnostics, competitors presently on the market count on stand-alone systems, which have the disadvantage that their use requires acoustic-phonetic expertise that is not available on a broader scope. PhonLab, on the opposite, exploits the central-lab service model that has proven vital in laboratory medicine, due to its capacity of providing cutting-edge diagnostic expertise across regional, social and economic boundaries. At market entry, PhonLab will have a competitive edge in the niche of neurogenic speech disorders, where the clinical research group of the hosting IPS already plays a leading role. From there, further market segments will be developed.
Achievements
At the end of the current ERC-PoC grant period, the technical components of PhonLab are all there, and a system architecture guaranteeing the scalability and medical device security of a market-adapted clinical service has been established in great detail. However, more time is still needed to implement all of the components to a point where a usable prototype can be launched and validated, because the substantial amendments of the already existing pilot system to make it secure, scalable and marketable had not been anticipated at project start.
All preparatory steps necessary to assess the usability and validity of PhonLab and to apply for medical device security upon completion of the technical work were taken. A cost-benefit analysis has revealed that PhonLab services can be offered at a price of less than € 15 per report, with a time-saving potential of ca. 25% of the time that is officially available for diagnostics. According to a business plan based on market and competitive analyses, break-even can be reached within four years after market entry.
Socioeconomic impact
The telediagnostic service to be offered by PhonLab after market entry is in full compliance with current and future developments in the digitization of health care (eHealth). The system’s socioeconomic benefit will be that it promotes the quality of health care provision for people who suffer from restrictions on participation in social life as a consequence of their speech impairment: It provides equal access to high-end diagnostic facilities, without requiring specific investment in expensive technology and across regional, social, and economic boundaries.
The objective of the ERC-Proof of Concept was to establish a commercial online service for the assessment of speech disorders. The service, named ‘PhonLab’, is planned to offer objective parameters over the web about the severity and specific pattern of voice and articulation impairment, given a sample of the patient’s speech. This is analogous to information delivered from a central clinical laboratory to a medical doctor about, for instance, cell changes, given a sample of the patient's tissue. In our case, the information includes diagnostic parameters concerning factors such as the degree of a patient's speech fluency or intelligibility obtained from an automatic analysis of the patient's acoustic speech signal provided via a web service. The results of this analysis are then returned to the user in the form of easily interpretable standard scores and graphics which provide clinicians with objective and sensitive data concerning treatment indication, treatment outcome, and the course of a speech disorder.
By providing high-end clinical and technological expertise independently of regional and social boundaries, PhonLab will enhance diagnostic standards for a large number of children and adults with speech impairments, initially in the German speaking countries, but with the longer-term perspective of a European and a world-wide service.
Resources.
PhonLab draws on techniques for processing child and adult speech acoustically that have been developed within the applicant's ERC project 'Sound Change and the Acquisition of Speech', as well as on research into the nature and treatment of speech disorders conducted at the host institute of the advanced ERC grant (IPS). The speech processing software developed at the IPS encompasses automatic speech segmentation tools, algorithms for the computation of all kinds of features describing a speaker’s voice and articulation, and tools required for the online recording of speech signals and the management of large speech databases. The clinical expertise available at the host institute includes experience with a wide range of clinical conditions and knowledge about the diagnostic validity and usability of acoustic speech parameters. Furthermore, a pilot system for web-based auditory assessment of speaker intelligibility has already been hosted and entertained at the IPS. The project’s main objective was to exploit these achievements and bring PhonLab to the point of initiating a start-up process towards commercialisation.
Beneficiaries.
The system, once established, will have extensive benefits to adult patients with neurological disorders of speech and language (stroke, Parkinson’s disease, Multiple sclerosis, etc.), children with acquired or developmental speech-language disorders, and children or adults with lesions of the speech organs, i.e. mouth, throat, larynx, and respiratory organs. The prevalence of these disorders with their devastating consequences for communication and participation amounts to an estimated 4 million patients in Europe.
The clinical caregivers of these patients, i.e. speech-language therapists working in clinics and private practices, will benefit from the cost-effectiveness and free availability of a service providing them quantitative and objective parameters they can use to improve their treatment planning and disease monitoring. Ultimately, the health care system and its stakeholders benefit from the cost-efficient provision of equal access to cutting-edge speech processing technologies across regional and economic boundaries.
The use of advanced speech technology and digital telematics for the clinical assessment of communication disorders is in line with efforts in Europe and the rest of the world to establish health telematics as a measure to enhance the efficiency of healthcare provision.
Challenges.
The following major obstacles were to be overcome in the development of PhonLab as a web-based service:
(1) The system must comply with EU regulations for Medical Device Security, including, among other things, data protection and quality management requirements.
(2) It must be scalable regarding the volume of orders per day/week, the scale of diagnostic parameters offered (e.g. voice, resonance, articulation, fluency etc.), the range of diagnostic indications (e.g. neurologic, developmental, ENT), the clients’ business forms (e.g. private practice, clinical unit), the specific documentation requirements of health providers and insurance companies, and the language in which assessment materials are presented.
(3) It must generate valid, reliable and customized diagnostic parameters in a most cost-efficient way, i.e. under conditions of almost complete automation.
Technical solution.
Building on experiences with an existing pilot system based on auditory instead of acoustic analyses, PhonLab was designed as a modular “Software-as-a-Service” (SaaS) system with the following components:
- A device for the flexible generation of assessment materials (words, sentences) from a large German corpus annotated for lexical parameters (word length and phonological transcript, syllabicity, word-type, lexical frequency),
- a browser-independent speech recording device which is designed to bypass the automatic speech processing routines implemented in the Capture-Device tools of most browsers,
- a transport system guaranteeing a secure transfer of speech samples from the end-user to the PhonLab-server by automatic “TLS-encryption”,
- a scalable and secure user management component allowing for a specification and management of various user types (patients, therapists, listeners, speech analysis experts, laboratory assistants, administrators),
- a tool chain encompassing programs for the automatic segmentation of speech samples and the extraction and analysis of local and global features characterizing a patient’s speech impairment, according to clients’ requests.
Privacy and medical device requirements.
Speech data acquisition, transport and storage, and the management of user access were designed to conform to the General Data Protection Regulation of the EU. Contracts of PhonLab with their clients (i.e. clinics, therapists) and of clients with their patients will be drafted according to legal stipulations, following the advice provided by an external data protection officer.
The requirements for certification as a medical device of security category IIa have been considered and documented on each step of the development taken so far. An application for certification will be submitted as soon as the system is ready for spin-off.
Market and competitors.
PhonLab’s target market at entry level encompasses speech and language disorders in children and adults in the German speaking countries, with an overall volume of more than 30 m Euros to date spent for diagnostics in Germany alone. Target users are speech-language therapists (28,000 in Germany) who work in ca. 10,000 units (clinics, private practices) and treat more than 850,000 patients per year. Further market development will expand on other language communities in Europe and world-wide as well as on fields of application outside the health sector, such as pronunciation training and assessment for migrants working in foreign language environments.
If acoustic speech analysis methods play a role at all in clinical diagnostics, competitors presently on the market count on stand-alone systems, which have the disadvantage that their use requires acoustic-phonetic expertise that is not available on a broader scope. PhonLab, on the opposite, exploits the central-lab service model that has proven vital in laboratory medicine, due to its capacity of providing cutting-edge diagnostic expertise across regional, social and economic boundaries. At market entry, PhonLab will have a competitive edge in the niche of neurogenic speech disorders, where the clinical research group of the hosting IPS already plays a leading role. From there, further market segments will be developed.
Achievements
At the end of the current ERC-PoC grant period, the technical components of PhonLab are all there, and a system architecture guaranteeing the scalability and medical device security of a market-adapted clinical service has been established in great detail. However, more time is still needed to implement all of the components to a point where a usable prototype can be launched and validated, because the substantial amendments of the already existing pilot system to make it secure, scalable and marketable had not been anticipated at project start.
All preparatory steps necessary to assess the usability and validity of PhonLab and to apply for medical device security upon completion of the technical work were taken. A cost-benefit analysis has revealed that PhonLab services can be offered at a price of less than € 15 per report, with a time-saving potential of ca. 25% of the time that is officially available for diagnostics. According to a business plan based on market and competitive analyses, break-even can be reached within four years after market entry.
Socioeconomic impact
The telediagnostic service to be offered by PhonLab after market entry is in full compliance with current and future developments in the digitization of health care (eHealth). The system’s socioeconomic benefit will be that it promotes the quality of health care provision for people who suffer from restrictions on participation in social life as a consequence of their speech impairment: It provides equal access to high-end diagnostic facilities, without requiring specific investment in expensive technology and across regional, social, and economic boundaries.