ARTIFICIAL INTELLIGENCE (AI) IS ALREADY ENCROACHING in many medical specialties and may render some obsolete.

This issue arose when my hospital’s radiology medical director explained that she had challenges recruiting young radiologists.

And a central culprit? Artificial intelligence (AI) and convoluted neural networks.

Machine learning approaches are targeting image-intensive specialties, including:

• radiology
• dermatology
• pathology

Specialties such as emergency medicine, anesthesia, plastic surgery, and my own, radiation oncology, have troubles, given oversaturation and encroachment from non-specialists.

Shortages

I entered medicine to help people. But job stability is an added benefit to being a physician.

But young physician trainees can no longer assume the stability I expected.

The Association of American Medical Colleges (AAMC) projects a physician shortage of 38,000 to 124,000 by 2034.

The causes included artificial intelligence, shifting healthcare paradigms, and oversaturation.

Radiology

I want to focus on radiology, which centers on interpreting medical images.

If you have a chest X-ray, CT scan, or MRI, a doctor known as a radiologist will interpret the images.

But medical students see the handwriting on the wall, recognizing how AI will increasingly encroach on radiology roles. Some believe misinformation is influencing students’ avoidance of radiology.

AI and radiology

Did you know a machine may already be involved in reading your mammograms?

Or adding a second read (to the human radiologist) to your breast MRI?

Artificial intelligence (AI), with machine learning driving continuous improvement, is now integrated into my daily life as an oncologist.

AI is the ability of an algorithm or set of computer instructions to mimic human behavior. The program can improve as we feed it more images and human interpretations of images.

Advanced clinical AI solutions offer advanced capabilities like automated image analysis to help increase radiologists’ efficiency.

The machine processes images faster than my expert radiology friends, can improve quality control in interpreting images, and works tirelessly.

Don’t worry (yet): A human is still central to the process.

Pathology

My hospital’s pathologists are central to my oncology practice.

A pathologist is a physician who examines bodies and body tissues. This doctor also oversees laboratory testing.

Before I can guide a patient’s cancer management, I need a pathologist to tell me the cancer type.

We increasingly use sophisticated tests to discover targets (on a cancer cell) for management.

A pathologist is a medical healthcare provider who examines bodies and body tissues. They are also responsible for performing lab tests.

This clinician helps other healthcare providers reach diagnoses and is a vital treatment team member.

AI and pathology

Pathology is a discipline that is remarkably vulnerable to encroachment by artificial intelligence.

Artificial intelligence (AI) is rapidly fueling a fundamental transformation in pathology practice.

My colleagues currently use artificial intelligence to help them interpret cancer cells.

For example, AI can help determine if a cancer cell over-expresses HER-2. This cell surface protein causes breast cancer cells to grow quickly.

Without treatment targeting the HER-2 overexpression, the cancer is much more likely to metastasize (spread to distant sites in the body).

Pathology — The future

Two observations from MDLinx immediately caught my eye:

First, by the year 2030,

A multitude of AI algorithms will be seamlessly integrated into routine pathology practices, having the potential to supplant pathologists in specific tasks completely.

Second, pathology has recently emerged as the most prolifically researched specialty among 17 medical disciplines engaged in AI-related research.

I look forward to artificial intelligence leading to the automation of cancer cell grading, interpretation of the spread of cancer to regional nodes, and assessing prognosis markers.

My colleagues will be increasingly efficient in the laboratory, and quality should improve.

Dermatology

Dermatology relies heavily on visual inspections and images.

Artificial intelligence is making rapid advances in image recognition. Can we teach it to recognize entities such as skin melanoma (or other cancer types)?

AI-based Deep Convolutional Neural Networks (CNNs) have arrived. Imagine this:

One study showed artificial intelligence outperforms dermatologists to diagnose skin melanoma.

The AI outperformed all of the 157 dermatologists.

Thoughts

Artificial intelligence has arrived, bringing efficiency and accuracy to selected medical tasks in imaging, tissue interpretation, and skin lesion interpretation.

However, we still need radiologists, pathologists, and dermatologists to provide nuanced interpretations. And supervision of the artificial intelligence community.

For now.