In conversation with Utcarsh Agrawal, data scientist at Akaike Technologies

1) How is data analytics in healthcare reducing costs for hospitals?

With the advent of Electronic Health Records (EHRs), healthcare professionals (HCP) can not only improve patient care but also streamline tasks and reduce errors, ultimately resulting in substantial cost savings. It’s now become easier to forecast demand for the operating room and ICU, resulting in efficient scheduling and resource allocation.

Accurate staffing is crucial, as half of a hospital’s budget goes into labour costs. Data analytics empowers HCPs to proactively tackle staffing challenges by forecasting needs and considering factors like historical data, local trends, and seasonal infections, leading to efficient shift management and cost-effective hiring operations.

Through data analytics, HCPs can identify high-risk patients and take preventive measures to lower 30-day hospital readmissions, ultimately reducing hospital expenses. Additionally, predicting patient no-show appointments helps HCPs offer free slots to other patients, enhancing revenue and customer experience.

2) Can you share examples of AI/ML techniques that have improved any healthcare project you’ve worked on?

Akaike Technologies is actively exploring ways to improve the effectiveness of medicines. Traditionally, drug efficiency trials involve testing medications on real-life patients directly through injections, which is time-consuming and expensive. However, by leveraging AI, computer vision and data analytics, we can now predict how well a medicine can work on patients, saving valuable time and reducing risks associated with human trials.

Another exciting project we are working on is detecting heart rate and 20 other vital signs using a mobile device. This data can empower HCPs and nutritionists to offer personalized healthcare plans to patients based on real-time information. 

Analytical tools, however, are a real game-changer in supply chain cost management. HCPs can track crucial metrics and automate processes with these tools, which are super-efficient assistants.  

3) Congratulations on the recent publication of your paper in Springer Link. What are the findings of your latest research?

As one of the world’s leading causes of death, lung cancer is the subject of my most recent paper. In this study, I focused primarily on reducing death rates while improving patient outcomes through early detection.

    Along with the assistant professor of JSS Technical Education, Ms Swathi Prakash, I explored the potential of machine learning and deep learning techniques to analyze histopathological images and predict early-stage lung cancer. 

    However, one of our challenges was training deep neural networks from scratch using medical imaging data. Gathering a large number of labelled images proved difficult. So, to address this obstacle, we delved into the world of transfer learning, fine-tuning pre-trained models to suit our needs. 

    Our study used the pre-trained EfficientNet-B0 model, which we fine-tuned explicitly for lung cancer detection. The results were impressive, with the designed model achieving high accuracy rates of 99.15% on the training set, 99.14% on the validation set, and 98.67% on the test set.

    Our research contributes significantly to lung cancer detection using deep neural networks, highlighting the potential of transfer learning in medical image analysis. This technology can play a significant role in reducing mortality rates associated with this devastating disease by enabling early and accurate diagnoses.