AI in the Medical Field

AI can be found throughout clinics and hospitals working alongside doctors, nurses, and specialists. Automated systems help flag risk factors, identify likely conditions, and assist medical professionals with other vital forms of diagnoses and treatment. Our resident brand journalist Leah Moore spoke to our in-house AI expert Cathy Feng to learn more about the state of AI in the medical field.


How can machine learning and deep learning help the medical field?

Machine learning and deep learning can help on multiple fronts. If we talk about computer vision, a type of AI that processes images, there are already many applications.

In medical departments such as radiology, ophthalmology, and neurology, doctors perform diagnoses based on MRI scans, CT scans, X-Rays, photographic images of tissue samples, and sometimes drawings of patients. Computer vision could help here to detect irregularities with quite high accuracy, sometimes even catching a diagnosis that doctors, pathologists, or neurologists miss.

While computer vision is mostly based on deep learning to process unstructured data, machine learning works more on structured data (e.g., age, gender, BMI, status of patients, etc.), and identifies potential hidden patterns.

A recent application developed in the UK by the Royal Marsden NHS Foundation Trust, the Institute of Cancer Research, and Imperial College London used a machine learning model to predict the likelihood of tumor regrowth based on an analysis of dozens of clinical data points and prognostic factors for patient. This model helped reduce human error and avoid potential bias in diagnoses to better identify possible diseases.

Traditionally, it has been the role of doctors to examine clinical data, scans, etc., and determine if they contain disease indicators. Although many medical professionals are experts in this field, they may miss subtle indicators from time to time. Some studies estimate human error at between 3-5 percent, while others warn that false positive rates prompted by a desire to be “more safe than sorry” can result in as many as 30 percent of lung cancer CT scan analyses being incorrect.


What are the benefits of automating test results?

Doctors make decisions or judgments using their domain knowledge, expertise, and experience, and sometimes these judgments can be subjective. Also, medical professionals must make decisions under pressure and challenging working conditions, which in some cases, incur human errors. AI can reduce subjectivity and eliminate human errors in addition to making such decisions more efficiently.

In addition, for tests that do not require special devices, like drawing tests used by neurologists, people may take them anywhere, anytime. This accessibility brings more benefits than simply reducing costs. It can help people identify diseases in earlier stages and slow disease progression.

Using AI creates a more level playing field for diagnosis, rather than relying on the varying abilities of doctors with different degrees of experience and expertise.

AI in the Medical Field - Clock Drawing Test

How have you used AI to detect Alzheimer’s?

CDT is a drawing test often used by neurologists for Alzheimer’s detection. In the test, patients are asked to sketch a clock face with numbers and hands pointing to a particular time. Then based on the drawing and various distortions in it, neurologists judge if the patient has symptoms of Alzheimer’s or not and decide the severity level.

We built a pipeline to simulate the human decision-making process in analyzing these sketches, incorporating computer vision, deep learning, machine learning techniques, and domain knowledge. Objects and features related to clock circles, numbers, clock hands, their corresponding positions, deviations, etc., were detected and extracted. Based on these features (50+), classification models and scoring engines were developed.

These engines are now plugged into a web platform that patients can access from anywhere. Patients can follow instructions to submit a photo of the clock drawing, then get a score for the prediction. The score ranges from 1 to 5, indicating whether the patient could potentially have Alzheimer’s and its severity level. Based on that score, people can decide if they should consult a doctor for treatment or if they are good for now.


What would you say to a doctor who doesn’t see the benefits of this technology?

These AI-driven diagnostic techniques have advanced to the point where doctors can use them to provide a “sanity check,” backing up an opinion the medic has already formed. They can provide reassurance in decision-making and ensure that fewer potential indicators of disease are missed.

AI-driven disease detection has the potential to be useful in many areas in addition to the examples already in play. Its developments should not be neglected.


What would you say to a doctor who is worried about AI replacing them?

Machines and humans are both important — machines bring efficiency, objectivity, and high accessibility, while humans bring knowledge, empathy, and flexibility to deal with the unknown.  All of these parts form a system that works well together. We do not see the possibility of doctors being replaced in the near future.


Talk to One of Our Experts

Get in touch today to find out about how Evalueserve can help you improve your processes, making you better, faster and more efficient.

Cathy Feng
Associate Vice President, ITC Forward Innovation & Data Analytics at Evalueserve Posts
Leah Moore
Brand Journalist Posts

Latest Posts