The Role of AI in Medical Diagnosis

Apr 3


Caroline Miller

Caroline Miller

  • Share this article on Facebook
  • Share this article on Twitter
  • Share this article on Linkedin

According to vast media reports, artificial intelligence (AI) technology is developing by leaps and bounds in each passing day with an influence on almost every aspect of our life. There are even views that many people’s work in the future may be replaced by AI. Regardless of the great transformations it might bring about, the active impact of AI on medical diagnosis is what we should applaud for.


AI diagnoses in skin cancer

A joint study by German,The Role of AI in Medical Diagnosis Articles French and American scientists found that artificial intelligence for the diagnosis of skin cancer surpasses dermatologists for the first time, helping to speed up the diagnosis process, helping patients fight cancer early and reducing the risk of normal moles being misdiagnosed as cancer.

According to the report, scientists used the Convolutional Neural Network (CNN) system to create the artificial intelligence system used in this study. CNN mimics the functioning of human brain nerve cells, enabling computers to recognize the immediate image like the human eye. The more images accumulated, the more accurate the judgment of artificial intelligence.

The researchers first entered more than 100,000 images, directed the AI system to distinguish between benign and potentially dangerous lesions, and then showed photos of 100 malignant melanomas or benign sputum to 58 doctors and artificial intelligence systems, comparing the ability to distinguish cancer of the two.

The doctors participating in the test came from 17 countries, more than half of them had over 5 years of medical experience, 19% had 2 to 5 years of experience, and another 29% had less than 2 years of experience. The results of the study showed that doctors correctly judged 86.6% of malignant tumors on average. If clinical data such as the patient’s age, gender, and suspicious wound location are provided, the diagnostic accuracy can be increased to 88.9%, but still less than 95% of artificial intelligence. In addition, doctors can correctly determine 76% of normal benign sputum, which is also lower than 83% of artificial intelligence.

The research team published a paper in Yearbook of Oncology, saying that the convolutional neural network system outperformed most dermatologists. Professor Heisler at the University of Heidelberg in Germany, the leading researcher, pointed out that artificial intelligence systems are less likely to misjudge melanoma, and at the same time are less likely to misjudge malignant tumors, exhibiting higher sensitivity than dermatologists.

The research team said that artificial intelligence will help speed up the diagnosis of skin cancer, allowing doctors to remove early cancer cells before they spread, and avoid unnecessary surgery for patients with benign sputum.

However, the researchers believe that artificial intelligence still can not completely replace the doctor, because the melanoma in the fingers, toes and scalp is difficult to record with photos, and some of the lesions are more special, which makes it more difficult for AI to identify. However, doctors can also help patients find out undetected suspicious wounds. Experts say that there is no way to replace clinical examinations at the current stage.

New algorithm for AI diagnosis can detect Alzheimer’s disease ten years in advance

Alzheimer disease (AD) is a central nervous system degenerative disease with insidious onset and chronic progression. It is the most common type of dementia in old age, mainly manifested as neurocognitive symptoms such as progressive memory impairment, cognitive dysfunction, personality change and language disorder, which seriously affect patients’ life. The etiology and pathogenesis of AD have not been elucidated. The characteristic pathological changes are extracellular senile plaques formed by amyloid beta deposition and neuronal fibrillation in the neuronal cells formed by hyperphosphorylation of tau, and neuronal loss with gliosis.

Recently, researchers at the University of Bari in Italy have developed a new algorithm that can detect tiny structural changes in the brain caused by disease. That is to say, AI can help achieve early detection of Alzheimer’s disease 10 years before the onset of symptoms.

The team trained their artificial intelligence through 67 MRI scans, 38 of which were from Alzheimer’s and 29 from healthy controls. The researchers divided the scan results into small areas and let their artificial intelligence analyze the connections between neurons. After the training was completed, they tested the algorithm by brain scanning 148 subjects. In addition to these tests, there were 48 scans of patients with this condition and 48 scans of people with mild cognitive impairment. Eventually, artificial intelligence was able to comprehensively detect Alzheimer’s disease.

According to reports, the research team’s artificial intelligence algorithm can diagnose 85% of Alzheimer’s disease, and the correct rate of detection of mild cognitive impairment can reach 84%, which makes it an effective tool for early diagnosis of potential Alzheimer’s disease. The significance of this study is that in the future it is possible to seek 10 years of valuable treatment and psychological preparation for Alzheimer’s patients. However, the researchers’ current test data is limited to the neuroimaging project database of the University of Southern California.