Technology in Oncology - Advancements and Challenges

Technology in Oncology - Advancements and Challenges

Healthcare Tech Outlook | Tuesday, January 11, 2022

The incorporation of technologies such as artificial intelligence (AI) is enabling radiologists and oncologists to diagnose, detect, and design treatment plans for patients with cancer.

FREMONT, CA: The healthcare industry has undergone massive transformations in the past few years to adapt to the demands of time and to incorporate the latest technological transformations. The latest technological transformations are helping to counter several conventional challenges in healthcare ranging from data organization, physician training, and even cancer care.

Technologies such as artificial intelligence (AI) are enabling radiologists and oncologists to diagnose, detect, and design treatment plans for patients with cancer. Cloud-based systems combined with AI allow the smart systems to sort and store massive data sets and organize them in a more useful manner. Virtual reality (VR) is also enabling additional surgical training.

Latest Developments against Cancer

AI is assisting healthcare and other industries alike. For instance, AI is being used to predict the severity of anxiety, depression, and sleep disturbance, which are linked to the decreased quality of life in patients with cancer.

Connected Care for Improved Outcomes

Connected care refers to the technology that can be used to promote cancer prevention, improve the experience of cancer care for care teams and patients, and foster progress in cancer research. The basic elements of connectivity are already present in various developed countries, including internet access, smartphone, and electronic health records (EHRs). The biological complexity of cancer and the requirement of deploying multispecialty teams make the use of technology particularly important for efficient communication and collection of information.

Cloud Storages for Cancer Data

The abundance of cancer-related data is vast for the human mind to sort and utilize reliably. A collection of 25 million tumor slides were shared by Memorial Sloan Kettering Cancer Center (MSK) which uses digitized data to train accurate programs. Real medical data is critical to developing efficient AI machines as-fabricated data can result in unintentional biases or failure.

Precision Oncology

Precision oncology research depends on the integrated analysis of various types of omics data, such as proteomics and transcriptomics. ML applications and algorithms are designed to automate various aspects of such a process at several levels of integration. It includes the integration of data at the model’s input level, predictions made by different ML algorithms and the combination of features derived from multiple data sources independently.

Latest examples relevant to precision oncology are cancer sub-typing derived out of the unsupervised learning of patterns noticed across copy number, point mutation, methylation, and RNA expression datasets. The prediction of several clinical outcomes of ovarian cancer patients using the combination of the epigenome, number aberration, and transcriptome datasets use a model that extracts predictive patterns from both unlabeled and labeled samples.


Despite the above-stated advances, there is still a requirement for expanding the range of predictive integration with respect to ML algorithms and data types. An important aspect is the prediction of patient outcomes or phenotypes depending on the integrative analysis of imaging, multi-omics, and other kinds of clinical data.

Moreover, the application of ML to several data sets depends on the availability of sufficiently matched, large, and carefully annotated datasets. Future applications will gain from ML strategies which are appropriate for dealing with relatively small datasets. Meaningfully linking multi-omics to imaging data will also rely on advances in processing and sample extraction technologies. Another challenge is the spatially-accurate matching of ex vivo data and in vivo matching.

On similar lines, deep learning relies on large collections of data. Apart from cancer research, typical DL systems are trained with millions of samples previously labeled by humans. Despite the continuous decrease in the cost of data generation, that includes molecular readouts through various technologies, access to large-high-quality datasets will be constrained. The challenge involves a relative lack of reference datasets containing and large truth.

Check Out : CIOReview | Medium

CIOReview | Pinterest

CIOReview | Facebook

CIOReview | Instagram

Weekly Brief