Photo by RetroSupply on Unsplash
How machine learning is being used to rank pathology results in precision medicine?
Machine learning is being used by medical professionals to improve precision medicine. There are many examples where machine learning is being used to treat patients and improve healthcare:
Cancer diagnosis and treatment: Machine learning is being used to improve cancer treatment for patients in healthcare. Cancer treatment involves developing new cancer diagnostic tools and such tools can have a profound impact on the patients being treated.
Rare disease diagnosis: Machine learning can also have a profound impact on rare disease diagnosis. For instance, a machine learning algorithm can be used to check for genetic mutations that cause cystic fibrosis.
Personalized medicine: Machine learning is revolutionizing the way individual patients receive treatment. Machine learning can be used to tailor treatment for specific patients. For example, it can identify whether a patient will likely receive treatment or not.
Drug Discovery: Machine learning-based algorithms can be utilized to develop new drugs. They can be used to make predictions about the efficacy and toxicity of new potential drugs.
Clinical decision support: Machine learning can also be used to design clinical decision support tools. Such clinical decision-support tools can be used to treat patients admitted in intensive healthcare units.
Medical Decision Making: Machine learning is also being applied to clinical decision-making.
Pathology results play a critical role in medical decision-making. Pathology results exist in various forms and depend on the type of test performed. Pathology results can be found in the following forms:
Written Reports: Pathology reports exist in the form of written reports. Such written reports depict the pathologist who performed the bloog tests and a detailed description of the findings of the past
Electronic Reports: Pathology reports in the form of electronic media are becoming increasingly common and can be used to diagnose a patient and can also be accessed from anywhere with an internet connection.
Images: Some pathology tests exist in the form of images like biopsies. Such tests are commonly used by pathologists to make a diagnosis and make a guess about their disease.
Numerical data: Some pathology tests, such as blood tests, produce numerical data. This data can be used to track a patient's progress over time or to compare their results to those of other patients. Such numerical data exist in the form of time series. Various forms of machine learning-based time series analysis can be done on such data to predict future patient treatments.
Audio or video data: Some pathology tests exist in the form of audio recordings which can be helpful to make a diagnosis.
The specific form of pathology results that a patient receives will depend on the type of test that was performed and the policies of the laboratory that processed the test.
Here is an example of a pathology result in the form of numerical data time series:
Date | Blood Glucose (mg/dL)
------- | --------
2023-09-01 | 100
2023-09-02 | 120
2023-09-03 | 140
2023-09-04 | 160
2023-09-05 | 180
The example shared above is a sample of blood glucose levels taken over a period of 5 days. The data is presented in chronological order and this can be used to track the progress of a patient over time. An increase in the blood glucose levels indicate a sign of diabetes. Here are some other examples of pathology results that can be represented as numerical data time series:
Heart rate
Blood pressure
Body temperature
White blood cell count
Hemoglobin A1c
Serum creatinine
Liver enzymes
Thyroid hormones
Machine learning algorithms can be applied to time series data in a variety of ways. Some common methods include:
Supervised learning: Given a set of labeled data, where each data point is associated with a label, a machine learning algorithm can be used to train a model A trained model then can be used to to predict the diagnosis for new blood glucose levels.
Unsupervised learning: Unsupervised machine learning models do not require a set of training data. The algorithm has to learn patterns existing in the data by itself and In unsupervised learning, the machine learning algorithm is not given any labeled data. The algorithm must learn to identify patterns in the data without any prior knowledge. For example, the machine learning algorithm could be given a set of blood glucose levels without any diagnoses.
I hope you found this article useful. In my next article, I will share how LSTM neural network can be used to predict blood glucose levels, given a set of time series data.