VADDOT: Vocal Analysis to Detect Diseases of Throat

Pharyngeal diseases, or throat diseases, are rampant- owing to pollution, allergies, and pathogens. The current diagnostic practices for such diseases are invasive, time-taking, and prone to subjectivity. Aiming to develop a patient-friendly mechanism for easy and efficient detection of throat diseases, reducing the time taken for diagnosis, VADDOT describes a throat disease identification system which utilises machine learning to detect diseases based on vocal parameters. It aims to provide an alternative, noninvasive way for the early detection of various throat diseases. VADDOT consists of Data Collection Module (DCM), Feature Learning Module (FLM), and Machine Learning Model (ML). DCM collects vocal samples by asking the concerned patients to record their voices while pronouncing various phonetics. FLM finds vocal parameters such as frequency, intensity, jitter, pitch, etc. from the voice. The Machine Learning Model works on Support Vector Machine trained by vocal parameter datasets of various throat diseases. Research in this field has shown promising results on diseases such as Parkinson's, Alzheimer's, and COVID-19. Since, the research is relatively new, datasets available to train the machine learning model are limited, causing difficulties in widening the spectrum of diseases that can be successfully detected through VADDOT. It is non-invasive, affordable, easy to use and reduces the diagnosis time. With further development and testing, the system has the potential to be widely used in healthcare facilities, enabling early detection and treatment of diseases, ultimately leading to better patient outcomes. It is simple and requires minimal training, making it an accessible tool for healthcare providers. Accuracy scores of 90.3% (Parkinson's Disease) and 97% (COVID-19) have been achieved so far by VADDOT.