My research goal is to integrate diverse computational methods to address biological questions related to complex diseases such as cancer. To this end, I have focused my research career on the application of machine learning techniques in precision medicine and healthcare. Machine learning is a wide-ranging technique that enables people to rethink how we integrate information, analyze data, and use the resulting insights to improve decision making.
I developed new methods for detecting biomarkers and critical molecular changes during cancer progression using expression data analysis . I also introduced a new algorithm for assigning types to interactions in gene regulatory networks inferred from microarray data .
In addition to omics data and network analysis, machine learning techniques interested me for my research stem from my goal of developing analytical methods. I demonstrate the utility of deep learning algorithms to classify different subtypes of cancer using digital pathology images. My framework (named CNN_Smoothie) and accompanying research have been named as a source by many other scientists in the field of machine learning .
This flowchart demonstrates the pipeline, which includes extracting data, training and evaluation of CNN algorithms, and prediction of various classes.
I extended my research on developing methodologies for analysis of large compendia that include not only pathological images, but also human blastocysts. I implemented an approach (called STORK) based on deep neural networks (DNNs) to select highest quality embryos for implantation after invitro fertilization (IVF) .
This flowchart illustrates the design and assessment of STORK.
Now I am working on developing an automated computational technique to distinguish aggressive prostate cancer from non-aggressive forms using MRI imaging data independent of biopsy. To standardize the evaluation and interpretation and reporting of prostate MRI, the prostate imaging reporting and data system (PI-RADS) is used for cancer detection. The PI-RADS score is subjective and relies on qualitative assessment of the physician interpretation and expertise. To address this issue, I introduce a deep learning based method (AI-biopsy) to distinguish aggressive prostate cancer from non-aggressive lesions based on Gleason score (pathology labels) using MRI imaging only .
I highlighted the regions of MRI images are being used by AI-biopsy for classification, and found that it focuses on the same regions that trained uro-radiolosts focus on.
My approach provides a reproducible way to assess cancer aggressiveness and uncovers potentially personalized strategies to reduce the number of unnecessary biopsies. Therefore, I will apply the method on other type of cancer such as Breast and Lung cancers.
My research addresses a fundamental need of humans for artificial intelligence (AI), which can apply intelligence to medical science. These methodological works described above are only part of my work. My collaborative efforts will direct me towards open problems to which I can apply my theoretical and mathematics skills in the development of useful and novel methodologies.