Accurate prediction of malignancy

Machine learning will predict the need for a mastectomy

Elizaveta Ivtushok, N+1

American researchers have applied machine learning to effectively predict the development of a malignant breast tumor. The algorithm was trained on patient biopsy data, and the accuracy of its prediction was 97.4 percent. The article was published in Radiology. 

The most effective way to treat a malignant breast tumor is mastectomy – surgical removal of the breast. However, if the tumor is benign, then even a qualified specialist cannot always predict with high accuracy how high the risk of developing cancer in the future is. The authors of the new work presented a machine learning algorithm that, based on breast tumor biopsy data, can accurately determine a benign tumor with an increased risk of developing breast cancer and, accordingly, identify the need for subsequent surgical intervention. 

To do this, the researchers collected 1,095 chest biopsy images obtained from 1,071 patients. In total, among the images, the researchers found 1006 tumors with a high risk of developing cancer, of which 115 led to the development of breast cancer. To train the algorithm, the scientists selected 671 images, and used the rest as a test sample. 

The algorithm was developed using the random forest machine learning method, a classifier that is based on the operation of many decision trees. Such a system stores the features of the data from the training sample and then applies them to correctly classify the data of the training sample.

The system developed by the researchers was able to correctly identify malignant tumors with a risk of subsequent development of breast cancer in 97.4 percent of cases. According to the authors, early diagnosis of pathology using machine learning methods could reduce the number of optional mastectomies by 30.6 percent.

Last December, scientists successfully applied machine learning and pigment distribution analysis for the first time to diagnose melanoma – you can read about this in our note. And here you can read about an algorithm that recognizes single (including cancer) cells.

Portal "Eternal youth" http://vechnayamolodost.ru Machine learning will predict the need for a mastectomy

Elizaveta Ivtushok, N+1

American researchers have applied machine learning to effectively predict the development of a malignant breast tumor. The algorithm was trained on patient biopsy data, and the accuracy of its prediction was 97.4 percent. The article was published in Radiology. 

The most effective way to treat a malignant breast tumor is mastectomy – surgical removal of the breast. However, if the tumor is benign, then even a qualified specialist cannot always predict with high accuracy how high the risk of developing cancer in the future is. The authors of the new work presented a machine learning algorithm that, based on breast tumor biopsy data, can accurately determine a benign tumor with an increased risk of developing breast cancer and, accordingly, identify the need for subsequent surgical intervention. 

To do this, the researchers collected 1,095 chest biopsy images obtained from 1,071 patients. In total, among the images, the researchers found 1006 tumors with a high risk of developing cancer, of which 115 led to the development of breast cancer. To train the algorithm, the scientists selected 671 images, and used the rest as a test sample. 

The algorithm was developed using the random forest machine learning method, a classifier that is based on the operation of many decision trees. Such a system stores the features of the data from the training sample and then applies them to correctly classify the data of the training sample.

The system developed by the researchers was able to correctly identify malignant tumors with a risk of subsequent development of breast cancer in 97.4 percent of cases. According to the authors, early diagnosis of pathology using machine learning methods could reduce the number of optional mastectomies by 30.6 percent.

Last December, scientists successfully applied machine learning and pigment distribution analysis for the first time to diagnose melanoma – you can read about this in our note. And here you can read about an algorithm that recognizes single (including cancer) cells.

Portal "Eternal youth" http://vechnayamolodost.ru