AI in Cancer Treatment

nvidias-candle-ai-to-help-fight-cancer-300x161

There have been approximately 14 million new cases of cancer in 2012 throughout the world which is further expected to rise by about 70% over the next two decades. Cancer is the second leading cause of death globally, and was responsible for 8.8 million deaths in 2015. In India alone, estimated number of people living with the disease is around 2.5 million, with every year over 7 lakh new cancer patients are being registered.

The scientists nowadays are more into development of deep learning or artificial intelligence based applications and services in order to increase efficiency and accuracy on a sustainable aspect.

A popular research area, both for scientists and medical practitioners around the world, right now is cancer detection Artificial Intelligence (AIs) that take a CT scan and determine whether or not the patient has cancer. They may even diagnose the type and characteristics of cancer. This can also lead to earlier cancer detection, as some systems may catch cancer long before a doctor can. It’s not perfect, but can be a huge help. Deep Learning plays a vital role in the early detection of cancer. It drops error rate for breast cancer diagnosis by 85%. In 2010, the total annual cost of cancer was estimated at around $1.6 trillion. But if you detect cancer early, your probability of survival is 10 times higher. Early detection can save not only billions of dollars but also countless lives. Deep learning has also shown capabilities in achieving higher diagnostic accuracy results in comparison to many domain experts.

So we build a new system. This one takes as features data about a person’s lifestyle and any symptoms they may be experiencing and any other health/environment/genetic data we have. It aims to predict a person’s risk of getting cancer. Those with high risk are then advised to get semi-regular CT scans.

So now when you go for your regular check-ups, you answer a few questions that get fed into a model and it determines if you are at risk for cancer. If so, it recommends a CT scan. Another model then examines that CT scan. It may diagnose you with cancer. Then what? Well, we probably feed the same scan and maybe a second scan through a host of other models just to double-check. If they come back positive as well, we need a treatment plan. What can AI do for you here?

Just ask IBM’s Watson. It is already being used to develop treatment plans for cancer patients. Watson reads medical journals and uses that information of the patient to recommend a diagnosis and treatment options. An AI for medical diagnoses can be like a search engine, but instead of typing in “cute cats in big hats” like you usually do, the query is provide all the information it has on a particular patient. It then selects all the relevant research or diagnoses or treatments it is aware of, discarding research that it deems irrelevant.

Finally, it ranks the relevant information based on relevance, likelihood of success,side-effects, costs, etc.This approach expands on a doctor’s expertise and saves time, allowing for faster, cheaper, more effective healthcare.

The actual treatment may not be any cheaper, but patients will pay less for the doctors’ time.

Applications of Deep Learning in Oncology Gene expression is the method by which the “genetic code” – the nucleotide sequence of a gene is used to direct protein synthesis and produce the structures of a particular cell. It plays an important role in Cancer detection. However, Gene expression data is very complex due to its high dimensionality, making it challenging to use such data for cancer detection.

Researchers from Oregon State University were able to use deep learning for the extraction of meaningful features from gene expression data, which in turn enabled the classification of breast cancer cells.

The technology was used to extract genes considered useful for cancer prediction and for the detection of breast cancer.

Cancer Classification with Deep Neural Networks Convolutional neural network are a category of artificial neural networks that is used to analyze visual images. Convolutional neural network (CNN) has achieved performance on par with all experts in classifying skin cancer. A single CNN trained end-to- end can classify skin lesions from images by using pixels and disease labels as inputs. This has kept the diagnosis reach limited not only to clinic but also extended to outside the clinic and into various service-based apps.

Tumor Segmentation

Deep learning is used for segmenting brain tumors in MR images and has given more stable results in comparison to manually segmentation of the brain tumors done by physicians which can have motion and vision errors. Deep learning can differentiate between benign and malignant breast tumors by using ultrasound shear-wave elastography (SWE). It has given more than 93% accurate results on the elastogram images of around than 200 patients.

Histopathologic Cancer Diagnosis

Diagnosis and grading of cancer has become increasingly complex due to increase in number of cases of cancer and patient specific diagnosis options. Pathologists go through a large number of slides in order to diagnose. Even there is an increase in the number of quantitative parameters pathologists considered nowadays to extract meaningful results (e.g. lengths, surface areas, mitotic counts). In that case deep learning can improve the efficiency of histopathologic slide analysis by increasing the objectivity of diagnoses and reduce the workload for pathologists. Therefore it can improve the efficiency of prostate cancer diagnosis and breast cancer staging.

Tracking Tumor Development

Deep learning can be used to measure the size of tumors and detect new metastases that might be over looked in manual diagnosis process. The Deep learning algorithm is developed by reading more and more CT and MRI scans which will help to improve its accuracy. The algorithm has attained localization score of 89%, in comparison to 73% accuracy rate achieved by pathologists.

Prognosis Detection

By using Deep learning, prediction model is developed which is used for the prognosis of patients suffering from gastric cancer and undergoing treatment (i.e. gastrectomy). It showed higher survival predictive powers as compared to other prediction models.

Conclusion:

The use of deep learning in oncology increases the chances of survival. Machines are helping researchers to find a coveted cure and prevention methods for the development of cancer as early cancer detection and prognosis can save the patient. It is the important healthcare areas where deep learning can be applied to improve the scenario of cancer treatment.

Sources:

[1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5661078/

[2] https://www.sciencedirect.com/science/article/pii/S1361841516300330

[3]https://www.forbes.com/sites/bernardmarr/2017/05/16/how-ai- and-deep- learning-is- now-used- to-diagnose-cancer/#486ca786c783

—- By Sonika Aneja and Aditya Gupta

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s