Welcome to Gravitas

5 Ways AI can be used in the Area of Cancer Research

With the emergence of newer technologies every day, the world is becoming a smarter place to live in. In fact, potentially every sector is undergoing a sort of pleasant transformation ever since the wave of Artificial Intelligence has hit the globe.

But what is AI?

For those who are technology amateurs, artificial intelligence, widely recognised as AI, is centred on developing intelligent machines that resemble human intellect in terms of reasoning, learning, and decision-making.

However, AI has so far been associated with simplifying standard tasks such as process automation, analysing & understanding consumer behaviour, or personalising experiences across sectors like banking, retail, and even education.But what’s interesting is the latest review report by Drug Discovery Today which suggests that AI can also be immensely helpful when it comes to the area of cancer research.

The report, authored by Vaishali Y. Londhe and Bhavya Bhasin, outlines in detail how AI’s applications are redefining the way oncologists approach the management of this big disease along with the 5 major areas where AI is making an impact. Let’s decode these areas:

1. Diagnosis of Metastases

The diagnosis of skin cancer is a long-drawn process that involves clinical screening, dermascopic analysis, biopsy, as well as histopathological analysis. But the use of the newest AI advances in this domain can make things faster, says the report authors.

Esteva et al. had conducted a study in the year 2017 which was published later in Nature. The study utilised a total of 129,450 skin cancer clinical images for training the convoluted neural network (also called CNN) on the ways to identify as well as classify cancers. Here, AI was successfully used for detecting malignancies.

Another study also produced fruitful results. A research team from the Oregon State University made use of the deep learning technology to extract details from gene expression data. It not only helped them categorise breast cancer cell types but also unveil new biomarkers for detecting breast cancer.

2. Tumour Segmentation

Tumour volume analysis is the very next step once the diagnosis is done with. Here, we must note that conventional methods that radiologists utilise, for instance Response Evaluation Criteria In Solid tumours, take a lot of time. Moreover, their accuracy level is merely 50 per cent.

For a greater standard of accuracy, scientists have utilised CNNs for segmenting optic path gliomas, brain tumours, and liver tumours. Consider the case where a study was conducted on liver cancer. CNNs were used by the team for liver tumour segmentation in the follow up CTs, baseline CT scan inputs, CT scan along with follow up scan definitions into CNN so that automated segmentation could be achieved successfully.

According to the paper authors, CNNs have an edge over the semiautomatic methods. That’s because it eliminates the requirement to custom make handcrafted features as it can identify the features automatically.

3. Precision Histology Application

As per the authors, precision histology has revolutionised histomorphology significantly. Precision histology can be understood as one of the deep learning forms.

Even though pathology and diagnostics have relied on the accurate interpretation of H&E stained slides, we can’t ignore the fact that this process can be relatively unreliable as well as time-consuming. However, DNN, short for Deep Neural Network, can boost the process thanks to its capability of utilising algorithms.

Interestingly enough, DNNs have already found its utility in skin lesion analysis. Its results are accurate as well, similar to the results one gets from the expert dermatologist. From breaking down the images into the form of pixels to combining them for developing reproducible characteristics which help generate a certain diagnostic pattern type.

Both Vaishali and Bhavya feels that DNNs would be able to provide analysis with increased accuracy based on the H&E slides on the grounds of further developments in whole-slide and high-throughput scanning technologies. Additionally, it would make way for developing newer biological data pool that would take precision oncology even further.

4. Tumour Development Tracking

Another way deep learning concept has been utilised is for tracking the development of a tumour. A research team at Germany’s Fraunhofer Institute for Medical Image Computing developed a specific deep learning model which can get itself updated and be more accurate day by day as it comes across more numbers of MRIs and CTs for readings. Also, the software even facilitates image comparison in a hassle-free way so that the tumour development can be accurately tracked between the clinic visits of a particular patient.

This approach would immensely assist in the detection of bone cancer, rib cancer, as well as spine cancer because such types of cancers are frequently missed, states the authors.

5. Cancer Stage Assessments

It is utmost important to evaluate any patient’s cancer stage with greater accuracy for prognosis. However, that can’t be attained with conventional assessment methods as they demonstrate a number of limitations, opine both the authors.

In order to overcome this scenario, researchers worked into the domain and developed a dedicated projection model which had deep learning at its base for predicting the chances of survival of a patient who had gone through the gastrectomy.

Based on the research, the authors state that the prognosis detection based on the deep learning tech displayed superior ability of prediction compared to the predictions done using the conventional Coz regression. They noted that the model based on deep learning could provide more precise and individualised risk-backed stratification.

To Wrap Up…

Malignancies have inherent complexities that make it greatly difficult to identify, map the progress, and early diagnosis of cancer. That’s why it becomes even more important to use state of the art concepts and technologies in cancer identification and treatment. However, as noted in the report, AI is transforming the cancer research domain slowly but steadily.

AI has all the potential and capabilities to overcome a majority of hurdles present in cancer identification, prognosis, growth, and treatment. But the transformational curve that AI brings to the table for cancer care seems promising to help boost the survival rates of patients.

Mukesh Kumar Sinha

Leave a Reply

Your email address will not be published. Required fields are marked *

Contact us

Contact us