Physical exercise had always been one of her daily “must-dos,” and Gina set out on her daily seven-mile walk with her walking buddy. Each day, they met for coffee, had a quick gab about what was new or exciting, and began the walk to the next town and back again.
Exercising as she did, one had to wonder why she had such firm, exercise-induced legs and calf muscles, but her stomach never seemed to go down. Her waistline was out of proportion to the rest of her body.
But Gina didn’t give it a second thought. She felt fine, was in a great relationship that would lead to marriage, and she was happy with her job. What could be wrong?
Today there was something wrong; pain in her stomach. Had she eaten something the night before, or was it that cocktail she tried at the birthday party for her fiance? Healthy as she was, it couldn’t have been anything; it was. Gina’s life was about to be turned upside down.
I didn’t see her for a month or more, and when I did, there was a decidedly different Gina. Her face was drawn, her mood didn’t have that “who cares” look, and her eyes pierced into mine. “I have cancer,” she said, trying not to cry.
Gina didn’t have a small cancerous growth. The tumors, the ones that made her stomach bulge, were numerous and spreading, jumping around her liver and her intestines, planting new tumors. The pain was the first indication she had. But her’s wasn’t the prime cancer killer. She did smoke, but lung cancer wasn’t part of her problem.
Attack on the lungs
The International Agency for Research on Cancer has estimated that, in both sexes, lung cancer is the most commonly diagnosed cancer and the leading cause of cancer death around the world. Detecting lung cancer has become a prime area of interest for the medical community. A new detection weapon had to be found to diagnose it earlier.
AI is providing more accurate cancer diagnoses and recognizing patterns through the interpretation of images and medical scans quicker. Computers are training in the detection of patterns following a cancer algorithms that enable the AI to become better at interpreting what the images present.
Researchers at Google and several medical centers have engaged in vigorous efforts to help pathologists read microscopic slides to diagnose cancer and to help ophthalmologists detect eye disease and persons with diabetes. “We have some of the biggest computers in the world,” said Dr. Daniel TSC, a project manager at Google. “We started wanting to push the boundaries of basic science to find interesting and cool applications to work on.” The “coolness” of cancer fueled interest in projecting AI as a prime hunter for cancer detection.
Lung cancer, which killed an estimated 160,000 persons in the US in 2018, is one of these prime targets of new AI technology. The American Lung Association has indicated that “approximately 541,000 Americans living today have been diagnosed with lung cancer at some point in their life.” During 2018, an estimated 234,030 new cases were expected to be diagnosed.
Primarily a disease of the elderly, 86% of those living with cancer in 2015 were 60 years of age or older.
Scoring better than thought
“Tested against 6,716 cases with known diagnoses, the system was 94 percent accurate. Pitted against six expert radiologists, when no prior scan was available, the deep learning model beat the doctors. It had fewer false positives and false negatives. When an earlier scan was available, the system and the doctors were neck and neck.”
Utilizing existing patient data without computed tomography imaging, one study had exciting results. “…our model outperformed all six radiologists with absolute reductions of 11% in false positives and 5% in false negatives. Where prior computed tomography imaging was available, the model performance was on-par with the same radiologists.” Again, its accuracy was proving worthwhile and promising.
AI technology is fast pulling ahead
Lung cancer is only one type of disease that is targeted by AI technology. Often, even when a patient is regularly screened for breast cancer via a mammogram, the condition is not detected at an early stage.
A professor at the Massachusetts Inst. of Technology, Regina Barzilay, had her own experience with a breast cancer diagnosis cancer, which redirected her attention to sophisticated technology for the detection of the disease.
In an interview, she noted that “Going through it, I realize that today we have more sophisticated technology to select your shoes on Amazon than to adjust treatments for cancer patients. I really wanted to make sure that the expertise we have would be used for helping people.”
Her experience led to a new collaboration with the Massachusetts General Hospital, which is now utilizing artificial intelligence and machine learning to improve the diagnosis and treatment of cancers. The questions which they are proposing deal with early detection in mammograms. As Barzilay said, “It clearly didn’t just appear” and hadn’t been detected even though it was present during her regular exams.
One new machine learning tactic in the diagnosis of breast cancer is PathAL, formed by a Harvard pathologist, a computer scientist, and a collaboration of biotech firms and scientists. Other efforts at early disease detection outside the area of cancer include Deep Patient, which scours millions of patient records to make health predictions. This approach can identify diseases as diverse as schizophrenia, cancer, and diabetes.
An algorithm targeting cervical cancer, too, has been developed by a collaboration of the National Cancer Institute and Global Good. More than 9,400 women participated in the 18-year-long study and utilized more than 60,000 cervical images. The AI outperformed human visual inspection and identified precancerous cells earlier. Enter another major tech firm in breast cancer detection.
Google’s deep learning tool was 99% accurate at breast cancer detection in two studies, which were performed using its algorithm. It is expected that this preliminary research will be utilized to further expand deep learning into computer diagnosis and care in the future.
The major areas for cancer research
No cancer is insignificant, but the early detection and treatment of some may be more straightforward than others. What needs to be addressed by AI in this search?
“The uniqueness of cancers makes the mapping of their progression and early diagnosis difficult,” two researchers from the Shobhaben Pratapbhai Patel School of Pharmacy & Technology Management in Mumbai, India, wrote. “Deep learning has been applied successfully to areas that were previously difficult to understand and is setting new standards of cancer care.” They suggest AI has a significant role in:
Diagnosing metastases
Segmenting tumors
Applying precision histology
Tracking tumor development
Assessing stages of cancer
AI initiatives are growing like mushrooms, and the more programs we have, the better it is for everyone. One additional aspect of machine learning, of course, is that the programs teach each other and immediately benefit from any changes they perceive in the data. What scientist wouldn’t want that?