Automating the analysis of images, which depict lung, would allow to achieve accurate classification of neoplastic disease treatment and projections for the forthcoming period. Doctors do not detract from the biopsy data, but new studies suggest - held computer estimates exclude subjectivity and give an accurate diagnosis immediately.
Computers can be taught pathology assessment - this say researchers at Stanford University School of Medicine. Automated mechanism eliminates doubts and external influence factors. The machine detects lung cancer, detailing the extent step of treating a disease prognosis.
Dr. Michael Snyder emphasized that the evaluation of pathology today is highly subjective. If you come together with two professional physician in assessing one slide they can give quite different figures. Equipment, the operation of which is based on complex quantitative measurements, the results will improve diagnosis and therapy.
At the moment, the experiment is based on the evaluation of patients with lung cancer pictures, but in the future the researchers plan to use the program for all types of malignancies. New method reveals the molecular mechanisms of tumor development.
For many years, cancer was evaluated by light microscopy. It is not always distinguished sophisticated equipment subtypes of cancer from one another. For example, had difficulties in assessing adenokartsenomy and squamous cell carcinoma - they are difficult to distinguish in the pictures. The extent of disease was evaluated subjectively by forming deviated patient prognosis.
To create a program to use 2186 images borrowed from the project "The Cancer Genome Atlas." Each image contains information about the extent, stage of disease, the patient's life expectancy after treatment. The computer has the ability to evaluate more specific characteristics than people. This mesh size, shape and texture of the nuclei of cells, the relationship between them in the space. There are over 10,000 individual traits, which makes the diagnosis equipment.