Stick out your tongue

Physicians often ask their patients to "Please stick out your tongue". The tongue can betray signs of illness, which combined with other symptoms such as a cough, fever, presence of jaundice, headache or bowel habits, can help the physician offer a diagnosis. For people in remote areas who do not have ready access to a physician, a new diagnostic system is reported in the International Journal of Biomedical Engineering and Technology that works to combine the soft inputs of described symptoms with a digital analysis of an image of the patient's tongue.

Karthik Ramamurthy of the Department of Information Technology, Rajalakshmi Engineering College, in Chennai, R. Menaka, Siddharth Kulkarni and Rahul Deshpande of the School of Electronics Engineering, at VIT University, India, and colleagues, have trained a neural network that can take soft inputs such as standard questions about symptoms and a digitized image of the patient's tongue and offer a likely diagnosis so that professional healthcare might then be sought if needed. The digitized images of the patient's tongue reveal discoloration, engorgement, texture and other factors that might be linked to illness.

Smoothness and "beefiness" might reveal vitamin B12, iron, or folate deficiency, and anemia. Black discoloration could be indicative of fungal overgrowth in HIV patients or prolonged antibiotic use. Longitudinal furrows on the tongue are associated with syphilis. Ulcers may indicate the presence of Crohn's disease or colitis and various other conditions. The team's automated diagnostic, however, utilizes the condition of the tongue in combination with other symptoms to identify whether a patient has any of various illnesses: common cold, flu, bronchitis, streptococcal throat infection, sinusitis, allergies, asthma, pulmonary edema, food poisoning and diverticulitis.

The current system allows diagnosis of fourteen distinct conditions but the team adds that they will be able to add eye images and use those as an additional hard input for their neural network and so extend its repertoire significantly.

###

Karthik, R., Menaka, R., Kulkarni, S. and Deshpande, R. (2014) 'Virtual doctor: an artificial medical diagnostic system based on hard and soft inputs', Int. J. Biomedical Engineering and Technology, Vol. 16, No. 4, pp.329-342.