Five Innovative AI-Driven Oncology Companies and Ten High-Impact Non-Imaging Applications Transforming Cancer Care

By Wang Jin and Huang Bangyu

Dr. Stefan Buttigieg believes that while we may not perceive the ongoing artificial intelligence revolution in our daily lives, these cutting-edge technologies are, in fact, driving medical innovation at an astonishing pace.

He predicts that artificial intelligence (AI) systems will be adopted in 90% of hospitals in the United States and 60% of hospitals and insurance companies worldwide, thereby providing more accessible, affordable, and higher-quality care to 70% of patients.

Moreover, healthcare applications in the AI market are expected to be rapidly adopted worldwide, with a projected compound annual growth rate (CAGR) of 42% by 2021. Frost & Sullivan even predicts that global revenues from these AI-driven healthcare applications will reach $6.7 billion by 2021.

What specific impacts will artificial intelligence have on advancing research into the treatment of diseases, such as cancer—a globally recognized medical challenge? Stefan Buttigieg explores the future of AI-driven cancer treatment and research. This article was translated by VCBeat (WeChat ID: vcbeat).

Oncology is a medical specialty dedicated to the diagnosis and treatment of cancer, encompassing three distinct subspecialties: medical oncology, radiation oncology, and surgical oncology. Artificial intelligence (AI) plays a significant role in each of these subspecialties.

How Does Artificial Intelligence Play a Role in the Research and Treatment of Complex Diseases Like Cancer?

Physicians at The University of Texas MD Anderson Cancer Center and the Palo Alto Medical Foundation in California have begun exploring the potential applications of artificial intelligence and big data in cancer treatment. They have proposed 14 application scenarios that could benefit cancer therapy research. AI researchers and clinicians have primarily categorized these scenarios into three main pathways to accelerate oncology research:

1. By further developing and integrating existing cancer registries, conducting analysis and interpretation from local to global levels to better understand cancer mechanisms (ranging from common to rare cancers). Large datasets provide a reliable evidence base, while artificial intelligence assists in the analysis;

2. Enhance global cancer treatment pathways by analyzing best practices and trends;

3. By vigorously promoting the implementation of cost-effectiveness trials.

In conventional medicine, we use clinical methods such as ultrasound, X-ray, computed tomography (CT), and magnetic resonance imaging (MRI) to detect cancer. However, these techniques are actually unable to fully identify many cancers.

Another approach involves analyzing microarray gene profiles. Although this method sounds complex, it requires only a minute amount of genetic material to detect cancer by assessing the expression levels of specific genes. The data generated from this genetic material can create massive datasets that must be analyzed. This analysis may take several hours to complete. However, AI can now rapidly perform this complex and time-consuming analysis.

Research initiated in 2001 demonstrates that artificial intelligence plays a pivotal role in this field. By 2017, researchers in neural networks were already classifying cancers using innovative techniques such as Gene Masking.

When scientists struggle to comprehend, let alone accurately predict, tumor behavior, the malignancy of tumors is fully exposed. Countless cancer patients and their families are grappling with the uncertainty of potential recurrence. When a few residual cancer cells survive initial treatment by chance or remain undetectable post-treatment due to their minuscule size, patients face the risk of cancer recurrence.

The collaboration between the Stanford Artificial Intelligence Laboratory and the Stanford University School of Medicine inspired the creation of TensorFlow and a database of 130,000 dermatological images, and was used to train TensorFlow algorithms to visually diagnose potential skin cancer. Most importantly, the results generated by this algorithm were consistent with those of a team of 21 dermatologists.

Artificial intelligence applications in oncology are becoming increasingly prevalent, with five companies currently warranting close attention. In recent years, the scope of application for image recognition technology—which identifies objects within images—has expanded rapidly, driven by advances in deep learning.

Enlitic, an emerging company headquartered in San Francisco, USA, has applied deep learning to the detection of malignant tumors such as cancer. The cancer detection rate of the system developed by the company is higher than that of radiologic technologists. Deep learning is a machine learning method that uses “deep neural networks” simulating the structure of the human brain. It can also be used in speech recognition and natural language processing, but its most significant achievements have been made in the field of image recognition.

Enlitic is using deep learning to automatically detect lung cancer nodules in chest CT images, with results showing higher accuracy than a panel of expert thoracic radiologists.

Insilico Medicine, founded in January 2014 and headquartered in Baltimore, leverages artificial intelligence for drug discovery, biomarker development, and aging research. The company applies deep learning algorithms to develop cancer therapeutics, such as immunotherapies. This August, Insilico Medicine entered into a strategic partnership with Bitfury, a global leader in comprehensive blockchain technology services, to jointly innovate solutions that integrate blockchain technology into healthcare applications.

Oncora Medical is providing predictive insights and risk analysis for radiation oncology. In doing so, the company is helping radiation oncologists make better decisions and leverage the diverse, valuable data they generate. This Philadelphia-based startup focuses on advancing cancer research and treatment, particularly in the field of radiation therapy.

Pathologists around the world diagnose cancer every day, a task that requires analyzing thousands of slides. If there were a simple method to help these experts filter out all normal slides and flag those requiring further investigation, it would significantly reduce their workload.

Proscia is a company that applies computational intelligence to oncologic pathology. It aggregates and curates tumor pathology data and images from around the world, applying them to frontline clinical cancer treatment. Proscia’s digital pathology platform enables pathologists and researchers to leverage pathological data from each slide. In the first half of this year, Proscia extended this cloud technology to oncologic pathology analysis, establishing a cloud-based database of tumor pathology slides that facilitates easy sharing among users worldwide.

The article “A New Era of Oncology with Artificial Intelligence” by Dr. Curioni-Fontecedro was published in ESMO Open, briefly explaining the current landscape.

Although this technology and research exist and are available for cancer care and research, they have not yet permeated the entire oncology community. Implementation in underserved areas will require physicians to secure funding and undergo training as oncology advances to the next level, thereby enabling the acquisition and deployment of these innovations.

The future of cancer research and treatment is bright. We look forward to a near future in which cancer will be addressed and resolved in a simple, seamless manner, offering patients hope and opportunities for recovery. Stefan Buttigieg also highlighted ten artificial intelligence application scenarios with significant implications for human healthcare.

Electronic medical records serve as a repository that aggregates all health data for patients. If policymakers need to allocate resources based on data from a specific year, this requires public health professionals, data scientists, and informaticians to collaborate in analyzing thousands of anonymized patient records according to specific clinical coding standards.

In the current environment dominated by paper-based clinical documentation, this is unattainable. Furthermore, the process typically takes several months, and many of the resulting outcomes lack scientific validity.

However, electronic health record (EHR) mining technology is not limited to applications in high-level decision-making contexts; it can also be used to directly improve patients’ medical experiences. For instance, in recruiting patients for clinical trials, EHR mining technology can facilitate patient matching. In other words, Patient Recruitment Systems (PRS) can directly identify eligible patients and offer them opportunities to participate in clinical trials.

Imagine this scenario: on a snowy day, you feel as if your nose is about to explode and you are struggling to breathe. You find it difficult to decide between seeing a doctor and staying home to take some medication while waiting for self-recovery. Finally, you make up your mind to see a doctor and call to schedule an appointment with your most trusted physician. However, you are informed that they cannot see you today due to car trouble; they can only offer some basic advice and reschedule the appointment for another time.

In this era of rapid advancement in artificial intelligence, this will soon become a thing of the past. You simply need to open your smartphone and interact with your most trusted advanced consultation bot—a specialized application that enables conversational interaction and has undergone scientific validation by a team of professional and experienced physicians. The advanced consultation bot can provide practical advice and even facilitate video calls in emergency situations.

The transition from paper-based clinical documents to electronic clinical records is not as straightforward as we might imagine. In fact, for many physicians, data entry remains a challenging aspect of their daily workflow. However, this pain point has spurred the formation of strategic partnerships, such as the collaboration between Nuance Communications and Epic, one of the world’s leading providers of electronic patient records.

This collaboration will integrate the artificial intelligence capabilities of Nuance’s AI-powered Computer-Assisted Physician Documentation (CAPD) tool into the Epic NoteReader module, thereby improving clinical documentation. By leveraging deep learning and natural language processing to analyze relevant patient records, the Nuance CAPD tool can highlight specific clinical indicators within electronic health records and alert physicians when data is missing or requires clarification.

Physicians across various specialties engage in repetitive and monotonous tasks in their respective roles, while artificial intelligence can provide convenience in specific specialties or for particular roles. For example, radiologists need to review countless radiological images every day.

Most emergency radiologists may interpret more than 200 imaging studies per day, and a single CT angiography of the lower extremities can comprise up to 3,000 images. This can lead to eye strain, and since radiologists are a scarce professional resource in many countries, it is crucial to streamline this image interpretation process.

Therefore, the IBM R&D team developed an image-guided information system through the “Medical Sieve” project, which filters critical clinical information related to patient diagnosis and treatment plans for physicians, thereby enhancing their clinical workflow efficiency.

Nowadays, there are multiple avenues for improving patients’ healthcare experience. One of the most critical factors in this improvement process is connecting with the right providers at the right time within the healthcare system, known as patient access solutions. Kyruus and PokitDok are leaders in this field.

Patients can schedule appointments, identify suitable physicians, and undergo blood tests using simpler language. Currently, several startups have launched comprehensive solutions that combine AI with physicians to elevate the patient care experience to a higher level.

It is crucial that personalized treatment regimens may yield long-term positive therapeutic outcomes. Furthermore, with the support of clinical decision-making systems, clinicians’ concerns about concurrent treatment-related effects (such as adverse drug reactions) can be alleviated. Some technology companies have already taken a leading position in this field.

Among these initiatives, IBM Watson Health’s oncology division collaborates with oncologists to provide evidence-based treatment recommendations for clinicians. This approach involves analyzing the meaning and context of both structured and unstructured data in clinical records and reports, which has been proven critical in selecting personalized treatment pathways, particularly due to its ability to simultaneously integrate patient records, external research, and other relevant data.

Although the growing demand for nurses and physicians is already a globally recognized healthcare issue, talent development in nursing and medical schools appears to remain inadequate.

This inspired Sense.ly to develop Molly, the world’s first AI-powered virtual nurse. Through innovative technology, Molly can emulate the care delivery methods required by patients. This technology plays a crucial role for patients with chronic diseases who require long-term management, personalized monitoring, and follow-up care.

However, this technology plays a significant role not only in supporting chronic disease management but also in scenarios where patients must strictly adhere to medical instructions within a short timeframe, such as directly observed therapy for tuberculosis or clinical trials. The U.S. National Institutes of Health recommends the AiCure application, which utilizes smartphone front-facing cameras and artificial intelligence, to visually monitor and confirm medication ingestion, thereby ensuring patient adherence to their prescribed regimens.

Neura AI, an artificial intelligence startup dedicated to combating diabetes, is providing algorithms for Medisafe’s innovative medication reminder app. Neura AI’s algorithms can help physicians gain insights into patients’ daily routines—such as analyzing their sleep and wake times—to enable Medisafe to notify patients of the optimal times to take their medications.

Genetics and genomics will also be profoundly impacted by artificial intelligence, thereby driving the development of personalized medicine. Personalized medicine refers to the provision of specific, tailored treatment plans based on a patient’s individual conditions.

Deep Genomics, a company that brings together the world’s leading experts, has developed a generation of computational technology capable of predicting DNA variants. In a February 2016 TED Talk titled “How to Read the Genome and Build a Human Being,” Riccardo Sabatini demonstrated how his team successfully used a vial of blood and specific machine learning techniques to predict an individual’s physical traits.

In addition, Verily Life Sciences, a subsidiary of Alphabet Inc. (the parent company of Google Inc.), is also undertaking several projects focused on precision medicine to address diseases such as Parkinson’s disease, multiple sclerosis, and cardiovascular disease.

Drug development is not only time-consuming but also extremely costly. The use of AI can not only accelerate the R&D process but also make it more cost-effective. Atomwise has already addressed this challenge by leveraging supercomputers. Through virtual screening, they have been able to repurpose existing safe drugs for the treatment of Ebola virus, identifying two compounds that reduce the infectivity of the Ebola virus.

Atomite’s technology accomplished in a single day what would take months or even years to resolve using traditional analytical methods.

Simulation also plays a crucial role in the new drug development process. The data analytics company Insilico Medicine has undertaken a special mission to extend human lifespan by leveraging artificial intelligence for drug discovery and aging research. In collaboration with academic institutions, pharmaceutical companies, and cosmetics firms, the company has identified over 827 drug-disease predictions and biomarkers. This simulation technology reduces the need for animal testing and human clinical trials.

The development of emerging technologies has also led to the increasing maturity of medical imaging technology. Verily Life Sciences has solidified its position as a leader in this field by developing machine learning solutions for diabetes-related eye diseases through retinal imaging.

Current projects include: smart contact lenses capable of detecting diabetes indicators, the Liftware Spoon designed for Parkinson’s disease patients, the Baseline Study aimed at building a comprehensive map of human health by collecting genetic and molecular data from populations, and health-monitoring wristbands.

Machine Learning and Diagnostic Radiology

Zebra Medical Technologies has launched Zebra Medical Vision, an application focused on leveraging advanced machine learning and medical imaging to aid in disease diagnosis. By automating the analysis of millions of real-time and retrospective imaging studies, the company is training computers to detect and diagnose critical medical conditions. Zebra Medical Vision believes that providing machine learning researchers with the necessary tools and datasets can accelerate the development of advanced clinical decision support tools and diagnostic solutions, thereby delivering better healthcare services worldwide.

Static and Dynamic Imaging

Enlitic is dedicated to leveraging deep learning to assist physicians in interpreting medical images and enabling faster, more accurate retrospective analysis through real-time clinical support. Other startups, such as Butterfly Network, are exploring alternative dynamic imaging modalities, such as ultrasound.

Anatomical Pathology

Medical imaging encompasses various types of images, with those related to anatomical pathology being the most critical. 3Scan addresses this challenge by transforming manual, analog, and qualitative practices into automated, digital, and quantitative medical science, thereby enhancing the accuracy and efficiency of anatomical pathology.

During the 2016 Rio Olympics, the Zika virus was well controlled throughout the event, a success closely tied to the significant role played by the epidemiological initiative A.I.M.E.

The company specializes in the study of infectious disease transmission. A.I.M.E’s platform provides users with precise geolocation data and forecasts for the next outbreak of an infectious disease, offering a three-month advance warning with an average prediction accuracy of 86.37%. Accompanying these predictions is the company’s fully customizable analytics platform, which enables users to analyze public health data by providing timelines, medical history records, and reports sourced from social media.

References:

https://blog.infermedica.com/10-ways-that-artificial-intelligence-is-shaking-up-healthcare-part-1/

https://blog.infermedica.com/10-ways-that-artificial-intelligence-is-shaking-up-healthcare-part-2/

https://blog.infermedica.com/how-ai-is-changing-cancer-care-and-research/