How HBKU is Leveraging AI to Tackle Diseases

How HBKU is Leveraging AI to Tackle Diseases

02 Oct 2022

CSE is actively working to develop and integrate novel AI-enabled tools and systems for the healthcare domain

In the third of a three-part series discussing the role of Artificial Intelligence (AI) within the healthcare sector, we speak to Dr. Tanvir Alam, Assistant Professor at the College of Science and Engineering, part of Hamad Bin Khalifa University, about his work in this field.

Is the healthcare industry making full use of artificial intelligence?

AI is a data-centric approach, so AI adoption in healthcare is a challenge due to the sensitivity of patient information and streamlining the new technology in an existing clinical setup. To fill this gap, the College of Science and Engineering (CSE) is actively working to develop and integrate novel AI-enabled tools and systems for the healthcare domain, for both infectious and non-infectious diseases.

Can you tell us more about this work?

We leveraged state-of-the-art AI-based techniques on the Qatar Biobank cohort to diagnose diseases and discover associated factors. We developed an AI-enabled system for the identification and stratification of obesity-related risk factors. The proposed machine learning model achieved over 90% accuracy, thereby outperforming the existing state-of-the-art models.

Our system was able to confirm known risk factors for obesity and diabetes , and identify potential novel risk factors linked to obesity-related comorbidities such as liver function and lipid profile.

Complex diseases such as diabetes and cardiovascular disease (CVD) often involve an interplay among demographic , lifestyle, and genetic profiles. Therefore, seamless integration of multi-faceted factors is required to understand the disease landscape. AI-enabled systems have the potential to integrate and pinpoint the root causes of diseases.

We have also integrated a multimodal dataset from Qatar Biobank to develop an AI-enabled system that can identify risk factors for CVD. The system can detect CVD onset with 93% accuracy.

Is this system effective for everyone?

We tested the system on both males and females and age-stratified groups (young adults, middle-agers, seniors) and it showed similar accuracy for all strata. We compared the performance against the clinically approved scales such as the Framingham scale and atherosclerotic cardiovascular disease (ASCVD) scale, and our system performed better than the existing scale in clinical setup for CVD diagnosis plans.

Tell us about your work involving retinal images

We have developed a seminal method for diabetes and CVD diagnosis based on retinal images and dual-energy x-ray absorptiometry (DXA) scans. The proposed AI-enabled system diagnoses CVDs using fused information from retinal images and DXA measurement. The method provides a fast, non-invasive, and low-cost mechanism to obtain a CVD diagnosis. Retinal images provide medical practitioners with a fast way to examine the nerves of the eyes and diagnose diseases such as diabetic retinopathy, hypertension, or arteriosclerosis. We are currently working on the technology transfer of the AI-enabled system from the laboratory environment to the clinical environment. To fulfill this step, we are developing an action research-based framework to evaluate our AI-based models in a clinical set-up involving medical practitioners, nurses, and researchers working together in local hospitals.

Were you involved in any projects in relation to COVID-19?

We developed a computational method to predict binding between human microRNAs and the novel Coronavirus. Based on our prediction method, we identified 14 high‐confidence binding and we found that these miRNAs are expressed in diverse cell types that are relevant to COVID‐19 pathogenesis. Therefore, controlling these bindings would provide insights into the ongoing quest for an effective RNA-based drug for COVID-19.

How can this work be utilized?

Leveraging state-of-the-art AI-enabled natural language processing techniques, we analyzed the largest collection of scientific evidence from multiple sources and were able to pinpoint a few drugs such as Remdesivir, Statins, Dexamethasone, and Ivermectin that could be considered as potential effective drugs to improve clinical status and lower mortality in patients hospitalized with COVID-19. Our system also highlighted that Hydroxychloroquine could not be considered an effective drug for COVID-19. The resulting knowledgebase ( ) is available as an open-source platform for the community to discover possible ways for the therapeutic treatment of COVID-19. This research work will also help stakeholders to explore existing drugs that are already known to be effective in other respiratory diseases, ultimately supporting the community to shorten the potential list of drugs against COVID-19 and saving a huge amount of time in this research direction.