AI in Healthcare: 4 Ways Speech Technology Can Benefit Healthcare

Transforming the way we give and receive care

As AI becomes more common in businesses and society, many are starting to see the potential for AI in healthcare, where it can reduce costs, enhance the quality of patient care, and streamline the day-to-day jobs of medical professionals. From speech recognition to healthcare chatbots, AI is proving to be a valuable tool within this sector.

Automated speech recognition (ASR) technology holds particular promise. When combined with natural language processing (NLP), speech technology can understand, interpret and generate human language and perform tasks such as translation, transcription, automatic summarization, topic segmentation, and much more.

The ability to perform these tasks allows AI models to support human workers by giving healthcare professionals more time to focus on personalized, face-to-face patient care. For example, a nurse could use speech recognition technology to create patient discharge notes or update electronic medical records (EMR), giving her extra shift time to interact directly with patients.

Research supports the rising interest and demand for speech recognition technology and AI in healthcare. Accenture predicts the AI health market could reach a staggering $6.6 billion by 2021, representing a compound annual growth rate of 40%.

How Speech Recognition Technology Can Help Healthcare

Many healthcare companies understand the benefits of speech recognition, using the technology to create innovative solutions for some of our most pressing modern challenges. 

ASR saves time and increases productivity 

Underpinned by natural language processing, speech technology is helping doctors save time and improve productivity by taking over some of the labor-intensive administrative tasks related to their profession.  

Voice recognition documentation tool, Dragon Medical, for example, delivers real-time transcription, making it easier for healthcare professionals to keep comprehensive records. Instead of doctors typing up lengthy notes to fill electronic reports, Dragon Medical accurately translates their voice notes and memos into a detailed clinical narrative, allowing them to spend more time with their patients. 

Amazon Transcribe Medical offers a similar service. This machine learning tool quickly and accurately creates transcriptions of medical consultations between doctors and patients. The transcribed text can be analyzed using natural language processing and used to submit electronic health reports.  

AI-powered transcription may not seem groundbreaking, but when you consider that a 2017 study by the University of Wisconsin and the American Medical Association showed that primary care physicians spend six hours a day entering records into electronic health record systems, you begin to realize how the smallest of automations can make the biggest of differences. 

ASR improves healthcare communication 

Leading secure enterprise messaging service, NetSfere, incorporated medical-specific speech recognition technology into their mobile messaging platform to improve healthcare communications.

NetSfere’s solution provides a secure, encrypted messaging experience that accurately understands complex medical jargon shared through voice command. According to Raúl Castanon-Martinez, Senior Analyst for Workforce Collaboration at 451 Research, this allows frontline workers to “provide contextually intelligent, personalized and predictive delivery of patient care.”

According to a press release issued by NetSfere:

As speech technology using mobile messaging becomes more widely adopted in the healthcare sector, consumer-grade speech services are adequate for day-to-day communication but lack the ability to recognize industry-specific vocabulary routinely used in the medical field – i.e. anatomical and surgical terms, procedures, diagnostic tests, ailments, and prescription drug names. Doctors, nurses, and other medical professionals can now communicate and consult with one another through the NetSfere application regarding patient diagnosis or test results, leveraging speech recognition with the immediacy of messaging in a secure and encrypted fashion.

ASR allows remote diagnosis and care

Speech recognition technology also makes it much easier for patients and healthcare professionals to search for information about COVID-19 symptoms.

Mayo Clinic has added a skill to Amazon Alexa (called “Answers on COVID-19”), which acts as a self-assessment tool to help patients access clear and precise information about symptoms and determine if they require further testing.

Alexa’s new skill is one example of how speech recognition technology can help healthcare professionals deliver support at scale and remotely. For most healthcare providers, expanding their remote offerings eases the strain on overloaded systems and provides better protection against airborne viruses.

Assists in monitoring health status of patients

Speech technology is also helping doctors communicate with patients in more human ways. South Korea has enforced a 14-day mandatory quarantine for all those who have recently travelled into the country and for those who have been in contact with a COVID-19-positive patient. 

Technology is helping South Korean healthcare workers monitor the health status of these potentially infected people. SK Telecom enabled its home AI assistant “Nugu” to make twice-a-day voice calls to those under quarantine, to enquire about symptoms and analyze answers to determine the seriousness of the symptoms displayed. 

Amazingly, people who interact with Nugu are not restricted to ‘yes’ or ‘no’ answers and can provide the AI-model with complex, elaborate answers, just as they would to a human professional. This use of speech technology aims to reduce the heavy load of healthcare workers while improving the accuracy and efficiency of monitoring. 

The Role of Chatbots in Healthcare

Chatbots are another AI tool that is playing an important role within the healthcare sector. As telemedicine and virtual consultations are becoming an accessible and viable alternative to in-person doctor’s office visits, the first interactions with a healthcare provider in these settings will most likely be through a chatbot. 

Using chatbots has several advantages to both the healthcare provider and patient. For one, chatbots can be used to give standard information to the patient, such as what to expect for an upcoming procedure, vital information about postoperative care or how to prepare for a routine check-up. This type of information can easily be standardized for chatbot use. 

Chatbots can take care of other routine, but time consuming tasks such as screening patients by checking symptoms and asking a series of diagnostic questions, and tasks that a medical assistant would do. These include scheduling appointments, sending appointment reminders, alarms for when to take medications and long-term health tracking of vitals such as blood pressure, blood sugar levels and heart rate. 

Another advantage of chatbots is that, using high-quality training data, they can be trained to very specific instances and uses, allowing healthcare providers to collect a high volume of relevant information. On the patient side, this high level of personalization ensures that patients feel they are being well-attended and can receive detailed, relevant information. 

One recent example of effective chatbot use comes from a study published by researchers at Cleveland Clinic. Chatbots were used to gather information from patients scheduled for routine colonoscopies in order to identify risk factors for Hereditary Colorectal Cancer (HCRC) syndromes before the exam. The chatbot was also set to educate patients about HCRC and to get consent for further genetic testing. Results of this study were positive, showing that chatbots can be an effective tool in improved genetic screening and diagnosis. 

How to Train Medical Chatbots

In order to train medical chatbots, it’s important to start with the basics: what are the situations in which you will be using the chatbot, and what types of interactions will it need to prepare for? Once you have defined the scope, finding the right training data is the next essential step. 

The type of data required depends on factors such as subject matter, language and how the patient will primarily interact with the chatbot. No matter the subject, training data must be high quality and validated by a large group of people. Finally, it’s important to continuously train your chatbot to ensure that it is constantly evolving in the same way that patients’ needs and questions will. 

DefinedData is an ideal solution for training chatbots. Our off-the-shelf datasets are high quality, annotated and validated. See more about DefinedData’s catalog of training data here.

The changing face of healthcare

As can be seen from the above examples, ASR is already transforming the way people give and receive healthcare. AI can reduce the workload of medical professionals in many ways, freeing them up to spend more time on quality face-to-face interactions. The technology can also provide healthcare professionals with readily available information that will provide them with the insights they need to make important decisions. By combining human intelligence with machine learning, medical professionals are able to provide more comprehensive, more accurate and more engaging healthcare to everyone, everywhere. 

For more information about DefinedData, our off-the-shelf, readily available datasets that can help train your ASR in the language and industry that you need, have a look at our catalog of datasets here.