Artificial intelligence is now being used in a number of industries, which includes medicine. In health care facilities across the nation, the technological algorithms make electronic medical records more efficient. Robotics are used for surgery. Now, AI has proven helpful in the field of medical diagnostics. Recent studies performed by various scientists learned that algorithms and artificial intelligence have the ability to increase the efficiency in identifying cancer.
Convolutional Neural Network
According to research published in the “Annals of Oncology,” scientists from France, Germany and the United States trained a program known as convolutional neural network or CNN to assist dermatologists in diagnosing skin cancer. The technology was named for its ability to mimic the connections and communication that take place between neurons in the brain. CNN has the ability to receive, analyze and store data and images. The AI then teaches itself to perform and continually improve performance. The University of Heidelberg, Germany’s Department of Dermatology professor Holder Haenssle explains that CNN functions similar to the brain of a youngster.
Training the program involved entering more than 100,000 images depicting malignant melanomas along with benign moles and other lesions. The images were magnified 10x in order to display cellular anomalies between the different lesions. Testing the program pit the technology against 58 dermatologists from 17 countries. The group had a wide range of clinical experience ranging from two years to having practiced for many years.
Hundreds of images were presented to the CNN program and to the group of specialists. The physicians and CNN were asked to determine if the images represented malignant or benign lesions. They were also asked to create a treatment plan based on their decision. When the testing concluded, the artificial intelligence program demonstrated a higher degree of accuracy in diagnosing skin cancers compared to the human practitioners. Diagnosticians identified 86.6 percent of the malignancies. On the other hand the computerized program was correct 95 percent of the time.
One month later, the physicians were shown the same images. Only this time, they were additionally provided with specific details of each patient. Their diagnostic accuracy increased to 88.9 percent. However, they were still not as accurate as CNN. Professor Haenssle explained that an increase in correct diagnoses would reduce the number of unnecessary surgical procedures. The success of using artificial intelligence for diagnostics may make the technology eventually useful in a wider range of clinical applications.