Abstract

Radiation Guided Oncology using IoT and Machine Learning

Author: Raghu Bala

Background: Cancer is among the leading causes of death with over 14 million new cases and 8.2 million cancer-related deaths worldwide. Researchers are finding new methods of treating the disease and we believe that IoT and Machine Learning could play a pivotal role in this process. Method: Oncology is the study of cancer with three major areas: medical, surgical, and radiation. Medical oncology focuses on the treatment of cancer using chemotherapy. Surgical oncology involves the removal of cancerous tissues and performing biopsies for the detection of cancer. Radiation oncology focuses on treating cancer with radiation therapy. Today, the primary use of dosimeters by healthcare workers is to measure their exposure levels to radiation. We postulate that cancer treatments can be made significantly more effective by equipping patients with an IoT-enabled smart dosimeter which performs real time readings and stores the readings in the cloud for further analysis. In order to substantiate this claim, we have developed a smart radiation dosimeter (SRD) device equipped with a solid state nuclear sensor and combined it with circuitry to collect precise time, date and location information. The SRD collects radiation data in real-time and stores it in the cloud . This data is then analyzed using the NetObjex IoT Platform which includes an Analytics and Rules engine. We believe this approach can improve treatment protocols in the following ways: • Reducing errors: Cancer treatments need to expose patients to a certain amount of radiation at each session. If there are errors, it could be injurious to the patient and pose a liability risk for the hospital. Using the real time alerts of the platform, we can notify healthcare personnel of any dosage errors immediately. • Guiding treatments: Patients respond at different rates to radiation due to a number of factors such as age, gender, type of cancer, stage of cancer, diet, genetics, and more. Tying our SRD readings in the cloud to other data such as, size of tumor, or cancer biomarkers (e.g.EGFR, KRAS) from clinical readings to a time-series will create a personalized patient profile. This profile will correlate the response rate of the patient to the radiation therapy. Machine learning algorithms can be employed to throttle radiation dosage Our methodology for guiding treatments through IoT and machine learning has a number of benefits. • Through a corpus of data drawn from similar patients with a similar profile, machine learning algorithms can be trained to find the most optimal and, personalized treatment protocol. • The algorithms can also be further enhanced through a real-time feedback loop from a patient’s response rate to a given dosage. Conclusion: We believe that SRD coupled with machine learning provide real-time insights and can impact treatment regimens in an impactful way. We have focused our research and technology on the treatment of cancer. Treatments can be highly targeted based on biomarkers of a patient combined with exhaustive pattern matching capabilities of machine learning and the real-time response rates of patients.

Co Author/Co-Investigator Names/Professional Title: Srini Pagidyala, Advisory Board Member, Hoang Thanh Tung and Dr Divakar Krishnareddy