Primary tabs

Cancer is a highly prevalent disease of the 21st century, for which no cure has yet been found. It is characterized by out-of-control growth by abnormal cells, which multiply fast and crowd out the normal cells. Although cancer starts in one organ of our body (breast, blood, brain, lungs, colon, etc.) which defines its type, it can spread to other body organs through a process called metastasis. Many factors affect the prevalence risk, including the speed that the cancer progresses and the individuals response to various treatments which can include chemotherapy, surgery and radiation. Also of importance is the type of cancer, specific characteristics (number of tumors, tumor shape and size, proliferation rate, location within an organ), family cancer genealogy, gender, age, and other medical conditions. One of the key factors in the treatment of cancer is its early detection, which in some cases (e.g. breast cancer) can lead to a few decades of life expectancy.

Medical Imaging for Cancer

Medical imaging based screening techniques have long been used for the early detection of cancer and the scientific community considers that they play a major role in the reduction in mortality for certain cancer types. Advantages of medical imaging include minimal or no invasiveness, access to internal body organs without tissue destruction and functionality over wide ranges of time and size scales of biological and pathological processes. Medical imaging actually plays an important role in many other cancer management phases, such as biopsy guidance, staging, prognosis and therapy planning. Some of the widely used techniques are: ultrasound, projection radiography (X-ray), fluoroscopy, magnetic resonance imaging (MRI), nuclear medicine functional imaging techniques (positron emission tomography, single-photon emission computed tomography, scintigraphy), X-ray computed tomography (CT).

 

 

Emerging imaging techniques, still to be validated for use in clinical practice, try to differentiate between healthy and unhealthy organ tissue by examining properties such as elasticity (elastography), pressure (tactile imaging) or optical absorption (photoacoustic imaging).

The multitude of medical image techniques involved and the number of images generated for the cancer management of just a single patient, along with the increased prevalence of cancer worldwide (e.g. every year only in the United States there are over 5 million new cases of skin cancer) means that physicians have very tight time-constraints for their assessment of each image. As a result, a multitude of artificial intelligence algorithms, especially machine learning ones, are used more and more in medical decision support systems, in order to help doctors, do their job better and faster.

Machine Learning for Cancer Screening

The main purpose of screening, especially in the case of high-prevalence cancer types (e.g. skin cancer, breast cancer), is to detect cancer before metastasis, while the tumor comprises the smallest possible number of cells. Automated classification of medical images as normal or abnormal, and subsequently of detected tumors as malignant or benign, remains a challenging task.

Recently, the researchers at Stanford University demonstrated the potential of deep convolutional neural networks (CNNs) algorithms for dermatologist-level classification of skin cancer, which is primarily diagnosed visually. The largest ever dataset, comprising 129,450 clinical images and consisting of 2,032 different diseases, was used to train and test a CNN algorithm. The result was a performance comparable with that of each of the 21 board-certified dermatologists when trying to identify both the most common and the deadliest skin cancer. And what might that mean for you and me? It aims to provide low-cost universal access to vital diagnostic care through a mobile application. When combined with the huge number of Smartphone subscriptions expected as soon as 2021 (over 6 billion), it is clear why worldwide investors consider AI an opportunity for future profits.

Infervision: a Chinese startup promoting AI for lung cancer screening

China is one of the countries with the highest prevalence of lung cancer, which is further expected to increase in the next few years due to highly polluted environments. When combined with the poor-quality healthcare system, this provides the ground for high mortality rates of lung cancer, due to its fast progression (spreading via blood vessels will set early on distant metastases) and detection at rather late stages.

Doctors using Infervision's AI powered CT (Infervision)

Infervision, a Beijing-based company, uses machine learning algorithms and computer vision methods to support lung cancer diagnosis. The product is promoted as a “second pair of eyes” for the radiologists and it can identify more than 20 different cardiothoracic lesions, and can be used in physical examinations to screen for lung cancer characteristics and to quickly locate lesions. Chen Kuan, the chief executive and founder of Infervision, identified the opportunity through one of his relative’s experiences with the lack of qualified medical expertise in a small rural hospital. For its product development and continuous update, the company is using training datasets collected through the digital infrastructure existing in China since 2003, and it is also collecting real-time data from over 20 hospitals in China.

Niramai: India’s Non-Invasive Risk Assessment with MAchine Intelligence

The Niramai start-up comes to revolutionize breast cancer screening in another country with poor quality healthcare services: India. It proposes a multi-patented solution, SMILE, using a high resolution thermal sensing device and a cloud hosted analytics solution to analyze thermal images. Big data analytics, artificial intelligence and machine learning are combined for reliable, early and accurate breast cancer screening. Early results, from data of 300 patients collected in two hospitals and one diagnostic center, demonstrate high accuracy, which remains to be validated in large-scale pilot studies.

The fast support of these startups with funds from important investors in their countries, demonstrates the clear market potential and large scale impact of artificial intelligence in general, and machine learning in particular. It is hoped that in the forthcoming years, healthcare management will include self-screening for various diseases, using our smart mobile devices and low-cost services offered globally by companies like Infervision and Niramai.

Top image: A dermatologist uses a dermatoscope, a type of handheld microscope, to look at skin. Computer scientists at Stanford have created an artificially intelligent diagnosis algorithm for skin cancer that matched the performance of board-certified dermatologists. (Image credit: Matt Young)

Otilia Kocsis's picture

Dr. Otilia Kocsis

Dr. Otilia Kocsis received her BSc degree in Physics from the University “Al.I. Cuza” Iasi (Romania) in 1996, the MSc degree in Medical Physics from the University of Patras (1998), and the PhD degree in Medical Physics from the University of Patras (2004). From November 2002 until February 2008 she was working as a consultant for speech-based systems integration at SingularLogic A.E. (Greece). At the same time she was actively involved in a number of IST R&D projects (e.g. GEMINI, AMIGO, PO...Read More

No comment

Leave a Response