Scientists from MIT and Dana-Farber Cancer Institute have developed A new method of solving the problem of identifying the origin of cancer in patient subset. This challenge arises when doctors are not able to indicate the source of cancer, which makes it difficult to determine the most appropriate treatment, because many anti -cancer drugs are adapted to specific types of cancer.
The newly developed approach uses machine learning and involves creating a computing model. This model can analyze the genetic sequence of about 400 genes and use this information to predict the source of a specific tumor in the body.
With this method, the team managed to accurately classified over 40 percent of tumors with unknown origin in a set of data containing about 900 patients. This breakthrough allowed an unusual 2.2-fold increase in the number of patients who can potentially be candidates for personalized, genomically treated treatment, based on the identified source of their cancer.
Inteae Moon, the main author of the study and graduate of electrical engineering and computer science in MIT, emphasized a significant discovery that the model can potentially help medical specialists in making treatment decisions, leading them to personalized therapies in patients with cancer with unknown primary origin.
Alexander Gusev, senior author of The Article and Medicine professor at Harvard Medical School and Dana-Farber Cancer Institute, emphasized the impact of this work, especially on people with cancer with unknown primary origin, which applies to about 3 to 5 percent of cancer patients.
Traditionally, the lack of knowledge about the basic place of cancer origin has made it difficult for doctors to provide targeted treatment methods. These treatment methods, adapted to specific types of cancer, are often more effective and have less side effects than generalized treatment prescribed for a wide spectrum of cancer.
The study methodology focused on the analysis of routinely collected genetic data from Dana-Farber. The data included genetic sequences of about 400 genes commonly mutated in cancer. Scientists have trained machine learning model using data from almost 30,000 patients with 22 known types of cancer. Then this model, called OnConPC, was tested to about 7,000 previously invisible tumors with known origin. This showed an accuracy rate of about 80 percent, which increased to about 95 percent for high self -confidence forecasts.
After these promising results, the model was used to a set of data about 900 tumors from people with cancer of unknown primary origin. The model successfully generated high -certain forecasts for 40 percent of these cases.
The model's forecasts were additionally approved by comparing them with the analysis of the mutation of the embryonic line in the subbraze of tumors. The model forecasts are often adapted to the type of cancer expected by these genetic mutations. In addition, the model forecasts were adapted to the time of patient survival and their response to treatment.
By enabling the identification of the source of cancer, scientists have effectively expanded the pool of patients who could take advantage of the already available treatment. Research was supported by various foundations, including the National Institutes of Health and Louis B. Mayer Foundation.
Going further, scientists are aimed at improving their model by including additional types of data, such as images of pathology and radiology to provide a more comprehensive forecast covering different data methods. This may allow the model not only to predict the types of cancer and patient results, but also potentially recommending optimal treatment strategies.
Scientists often use machine learning techniques to detect diseases showing their versatility and potential in the field. The QDATA team additionally illuminated the necessary role of artificial intelligence and machine learning in healthcare through various projects. Our team made significant progress in the development of medical diagnostics, especially in the field of detection of cervical spine fractures. Our innovative solution intricately combines advanced image processing methodologies and machine learning models to recognize cervical spine fractures from CT scans. This pioneering approach, based on the analysis of bone structure and density, enables precise location of fracture places. As a result, doctors can quickly and accurately diagnose fractures, preventing potential complications and optimization of patient care.
In order to thoroughly examine the projects and achievements of Qudat, read the Qudata AI/ML case studies, which shed further light on a breakthrough work carried out by our team.
