Couldn't find what you looking for?


Can you imagine a scenario where the doctors have been replaced by computer based software? At present the scene may appear to be from a sci-fi movie, but we are surely and steadily moving in that direction.

Classification Accuracy Higher by Computer Software than by the Doctors

In a study funded by a German Research Foundation Grant, scientists tried to test the accuracy of face classification software. They had expected the results obtained in diagnosing a disease to come close to the results achieved by medical experts. But to their surprise, they found that the classification accuracy achieved by computer software was higher than by the doctors.

The researchers, led by Dr. Harald Schneider from the Ludwig-Maximilians University in Germany, wrote a computer program that analyzed the facial features of people suffering from acromegaly, a rare hormone disorder. They then took frontal and side photographs of 57 patients of acromegaly of a similar age group and of Caucasian race, and grouped them into mild, moderate and severe cases depending upon the facial features. All the pictures were analyzed using computer similarity analysis. The patients were categorized into these three factions by three acromegaly experts and three general internists also, on the basis of these photos. The computer software outperformed the doctors in identifying the patients suffering from acromegaly, especially that of the mild category. The software recognized six out of ten patients correctly whereas the physicians were successful in only four out of ten cases.

The study has been published in the Journal of Clinical Endocrinology and Metabolism.

Face Classification Technology can be Helpful in Early Diagnosis of Acromegaly

Acromegaly is a rare disease caused by unregulated production of growth hormone from tumors of the pituitary gland called as adenomas. It is often diagnosed in middle age and if untreated, can result in multiple illnesses like type 2 diabetes, hypertension, increased risk of cardiovascular diseases and arthritis. It also increases the chances of developing colon polyps and may lead to premature death.

As the disease is insidious in its onset, there is a considerable delay between the onset of the symptoms and the diagnosis of the disease. The disease produces some characteristic facial changes like protrusion of the lower jaw, enlargement of the nasal bones and spacing out of the teeth. Face classification technology can be helpful in the early diagnosis of acromegaly so that the treatment can be started earlier.

However, the technology has to be perfected before it completely replaces humans. It diagnosed acromegaly incorrectly in over 9% people who didn’t have the disease, in the aforementioned study. The sample was too small and the study was limited to people of Caucasian race of a certain age group. More studies are needed which include all the variations of the disease and people of other races too. It has to correctly identify the people who are not suffering from the disease.

The physicians who took part in the study where handicapped by the fact that they did not see the actual patients. Patients of acromegaly have several other features like abnormal hands and feet. The clinicians also didn’t know anything about the symptoms of the patients. To diagnose a disease based on a person’s photo is entirely different from a proper clinical examination.

Therefore, it is too early to say that humans can be replaced by machines in diagnosing diseases. However, the study looks promising and may be representing what will happen in the near future.

  • Harald J. Schneider, Robert P. Kosilek, Manuel Günther, Josefine Roemmler, Günter K. Stalla, Caroline Sievers, Martin Reincke, Jochen Schopohl and Rolf P. Würtz. A Novel Approach to the Detection of Acromegaly: Accuracy of Diagnosis by Automatic Face Classification. The Journal of Clinical Endocrinology & Metabolism, doi:10.1210/jc.2011-0237
  • American Association of Clinical Endocrinologists Medical Guidelines for Clinical Practice for the Diagnosis and Treatment of Acromegaly. Endocrine Practice. May/June 2004, 10(3):213–225
  • Photo courtesy by Rofi on Flickr: