Application of Machine Learning techniques to health diagnostics

Application of Machine Learning techniques to health diagnostics

Since computers began to be used on a large scale in the late half of the 20th century, according to needs, algorithms and programs have been developed to model and analyze large amounts of data.  In the late twentieth century, it was popularized the machine learning techniques, or machine learning. These techniques consist of making decisions and performing tasks based on artificial intelligence. Learning occurs by determining a set of rules, generated through the analysis of a database, or even by previously established models. This makes it possible for the generated model to recognize patterns, in short, to make decisions for the execution of tasks . 

The machine learning can specify in deep learning, used for interpretation and manipulation of data, ranging from pictures, voice, speech, interpreting texts and others. This model constantly learns, without a possible and natural cultural intervention, nor does it present difficulties common to human learning. With this, it can unite all the knowledge obtained, or input , and process it in search of an answer, or output , at a speed that no human could achieve.

Use of machine learning in the medical field

These deep learning techniques have been used since the beginning of the 21st century in modern hospitals, which are equipped with devices that gather and share large amounts of data in information systems; these data are applied for clinical analysis using artificial intelligence. 

Currently, studies show that artificial intelligence techniques can be more accurate than human assessment for diagnostics in the medical field. Added to this, in urgent cases, which require a quick medical intervention, the use of deep learning for a light analysis of data is even more necessary, so that, in a few minutes, a measure can be taken to avoid potential sequelae and even the death of patients.

In addition to its wide applicability, a single model of precise deep learning can be used in a convincing way in various forms of diagnostics in the medical field. Currently in the identification of several medical complications, such as diabetic retinopathy, cardiovascular risks, breast lesions, melanoma lesions, retinal diseases (from ocular images), spine problems, among countless other examples, even acting in the prediction of diseases. You can go deeper by clicking here .

A palpable example that we can briefly mention here is in the detection of skin cancer. In its diagnosis we noticed what are called melanomas and non-melanomas. Although much less common, melanomas are responsible for the highest mortality rate from skin cancer. In the image below, we can see the segmentation of the images by algorithms, defining squamous cell carcinoma in A and basal cell carcinoma in B, these being the most common forms of skin cancer, while in C the malignant melanoma, the most lethal cancer of skin.

Changes in pigmentation are the characteristics most considered in detection. However, many times due to the way the image is registered, several factors must be taken into account in order not to confuse the detection algorithm, causing bad segmentations of the image, as is the case of lighting, environment and instrument used in the registration. 

Will artificial intelligence systems replace humans?

 With this, numerous questions are raised about the possibility of the application of biotechnology, combined with information systems and bioinformatics in the creation of intelligent machines, to replace the professional activity of doctors . This is already a reality in many countries, such as the United Kingdom and China. 

This occurs through a personalized and automated service, which shows a great advantage, since it is possible to process simultaneous data from countless patients, obtain information in the scientific literature on how to proceed in a clinical case, in addition to being able to process and interpret the result of the interaction between the drugs to be administered.

A possible argument that patients would prefer to see a doctor directly can be easily refuted in some cases. This is because, nowadays we are hostages of the time, making many patients prefer a precise and quick service and diagnosis to make an appointment with a doctor. This prevents the appearance of setbacks, in addition to obtaining reliable results, independent of human presence.

Another factor that could stimulate the acceptance of these robots for care would be in cases where the patient has an eventual shyness in telling symptoms or behaviors and habits to a doctor.

Allied to this, the use of machines for medical diagnostics will be of great importance in places of difficult access , or where there are few doctors, as is the case in the Amazon – where many people have very rare consultations during their lives. In China, machines are currently used, not only for diagnostics, but also for surgery, as was the case with the first dental implant without any human interference.

However, even with all these prerogatives, human doctors will always exist, even if in smaller numbers, since these will still be necessary to develop the machines, in addition to operating them. Added to this, critical and human thinking on the part of them will be essential in many cases , and they cannot assume their extinction as a function and profession. 

A possible counterpoint to using robots to perform diagnostics and medical functions is that there will be a gap between the patient and the knowledge of the disease to be faced. This is because, there could be no direct contact with the artificial system as it is with a human doctor, preventing many doubts to be resolved in a natural way (since speaking with Google’s voice recognition system, for example, will never be like talking to another human). In addition, it would be necessary for the population to accept it, in order to avoid currents of thought such as “this machine does not know anything, I will not follow the guidelines”. 

With this, it is possible to notice the great advance of science, in general, contributing to public health, as well as democratizing it with efficient and practical ways of diagnosing and monitoring patients, whether on site or at a distance. Being a great and promising area of ​​Biotechnology and Bioinformatics that still promises much to evolve and revolutionize medicine and other areas, and may even make some professionals obsolete.

References

  • Esteva A, Robicquet A, Ramsundar B, Kuleshov V, DePristo M, Chou K, Cui C, Corrado G, Thrun S, Dean J. A guide to deep learning in healthcare. Nat Med. 2019 Jan;25(1):24-29. doi: 10.1038/s41591-018-0316-z. Epub 2019 Jan 7. PMID: 30617335.
  • Wiens J, Shenoy ES. Machine Learning for Healthcare: On the Verge of a Major Shift in Healthcare Epidemiology. Clin Infect Dis. 2018 Jan 6;66(1):149-153. doi: 10.1093/cid/cix731. PMID: 29020316; PMCID: PMC5850539.
Share this