|
Getting your Trinity Audio player ready...
|
Computer scientists from Iraq and Australia have harnessed the power of artificial intelligence (AI) and digital imaging to create an innovative diagnostic tool, achieving a remarkable 98% accuracy in detecting facial palsy, while also determining the patient’s gender and age.
The researchers, affiliated with the Middle Technical University (MTU) in Baghdad and the University of South Australia (UniSA), believe that the tool has the potential to significantly reduce diagnostic errors associated with this prevalent and treatable neurological disorder.
Facial palsy, characterised by temporary muscle weakness or paralysis on one side of the face due to impairment of the facial nerve, affects approximately 1 in 60 individuals worldwide during their lifetime. While facial palsy is commonly caused by nerve impairment, less frequent causes include tumors, infections, or strokes. The challenge in diagnosing facial palsy lies in its ability to mimic other conditions, often leading to misdiagnosis. A 2020 paper estimates that misdiagnosis occurs in up to 20% of cases, underscoring the critical need for accurate diagnostic tools.
Published in BioMedInformatics, the researchers detailed their real-time detection system for facial palsy, leveraging a microcomputer, digital camera, and a deep learning algorithm. The team utilised a robust dataset comprising 26,000 images, with 19,000 representing normal facial conditions and 1,600 showcasing facial palsy. Employing advanced AI techniques, the researchers trained computer vision systems to recognise facial palsy, distinguishing it from healthy conditions.
To put their system to the test, the researchers captured images of 20 patients with varying degrees of facial palsy. The algorithm not only detected the condition in real time but also accurately identified the approximate age and gender of the patients. According to Professor Javaan Chahl, a remote sensing engineer at the University of South Australia, the system achieved an impressive 98% accuracy rate.
The significance of this AI-driven diagnostic tool extends beyond its accuracy. By using computer vision systems for facial palsy detection, the researchers believe that the tool has the potential to prevent misdiagnoses, ultimately saving time, effort, and costs for both patients and medical specialists.
Traditional methods of detection through visual examination are prone to inaccuracies due to the subtle presentation of facial palsy, which can be easily mistaken for other conditions. Early and accurate detection is crucial, as facial palsy may be indicative of underlying issues such as stroke, HIV infection, multiple sclerosis, Guillain-Barré syndrome, or Lyme disease.
The researchers highlight that individuals most at risk of developing facial palsy are typically aged between 30 and 45, pregnant women, diabetics, and those with a family history of the condition. Moreover, facial palsy tends to affect the left side of the face more frequently, although the condition often resolves spontaneously within six months.
The paper titled “Automatic Facial Palsy, Age and Gender Detection Using a Raspberry Pi” is authored by Ali Al-Naji from MTU and UniSA, Javaan Chahl from UniSA, Ali Saber Amsalan from MTU, and Ammar Yahya Daeef from MTU. This collaborative effort between Iraqi and Australian researchers underscores the global nature of scientific advancements and their potential to transform healthcare practices. As the field of AI continues to evolve, such interdisciplinary collaborations pave the way for innovative solutions that have a tangible impact on patient care and medical diagnostics.
