trial success rates , thus reducing the R & D burden in the pharmaceutical industry . It can be used to reshape key steps of clinical trial design towards increasing trial success rates .
Machine Learning ( ML ), specifically Deep Learning ( DL ), can identify patterns in large data sets like text , speech , or images . Natural Language Processing ( NLP ) can comprehend and connect written or spoken language content , and human-machine interfaces ( HMIs ) allow for a natural exchange of information between humans and computers . These features can be utilised to correlate diverse and extensive datasets such as electronic health records , medical literature , and trial databases for better patient-trial matching and recruitment before the trial starts . Additionally , these techniques can be used for continuous and automatic monitoring of patients during the trial , resulting in improved adherence control and more reliable and efficient endpoint assessment .
Diagnostic and Treatment
The utilisation of AI in the improvement of user experience through the monitoring of physiological signals has proven advantageous in the diagnostic and treatment domains . An exemplary illustration is observed in the context of oropharyngeal dysphagia treatment . Oropharyngeal dysphagia refers to a condition where the ability to swallow is impaired due to various neurological impairments ( for example , Parkinson ’ s disease and stroke ), respiratory disorders , head and neck cancers , and genetic syndromes . Inadequate management of this condition not only negatively impacts a patient ’ s quality of life but also gives rise to severe consequences such as dehydration , malnutrition , pneumonia , respiratory complications , and potentially fatal outcomes . The current method employed for monitoring dysphagia patients involves videofluoroscopy ( VFS ), which proves to be effective but necessitates the expertise of highly trained speech-language pathologists ( SLPs ) to analyse the swallowing function using high-density x-ray images . Additionally , the VFS procedure requires patients to ingest varying quantities of barium bolus to enable visibility during the examination , thereby exposing them to radiation . Moreover , this approach is economically unfeasible for individuals with limited mobility or those residing in rural areas without adequate resources . Consequently , the development of a secure , portable , and non-invasive technique for monitoring the swallowing process becomes imperative .
Researchers from UTAR have successfully developed an AI-assisted , self-powered , and adaptable throat sensor for monitoring laryngeal movement ( Lee et al . 2021 ). In contrast to conventional methods , the sensor , enhanced with AI capabilities , enables more intricate perception functions and can operate effectively in dynamic environments . By affixing the sensor to the throat , muscular movements can be captured through the electrical voltage output generated by pressure exerted on the skin . The AI-assisted sensor proficiently identifies vital signals by detecting vibrations in the throat caused by actions like swallowing , humming , nodding , and coughing , with an impressive accuracy rate of 95 %. Notably , the proposed sensing system showcases substantial potential for enhancing the intelligence and efficiency of smart healthcare systems .
In addition to its applications in monitoring , AI has found extensive utilisation in the field of rehabilitation ( Lee et al . 2021 ). This article introduced a novel smart glove that incorporates an object recognition feature through the implementation of a supervised machine learning algorithm known as the support vector machine ( SVM ). The smart glove was meticulously crafted , employing five resistive flex sensors that could be attached to each finger individually . A thorough characterisation of the flex sensor was conducted , ensuring optimal performance . To facilitate data processing and analysis , a microcontroller was employed to receive and transmit the sensor data to a PC , where machine learning and prediction took place . The SVM model underwent training using a dataset comprising 160 instances , ultimately achieving an impressive accuracy rate of 91.88 % in recognising three distinct objects with varying shapes .
Benefits of AI in Healthcare
One of the key benefits of AI in healthcare is improved efficiency and accuracy . AI algorithms
55