Cancer Screening

Cervical cancer is one of the most common cancers affecting women worldwide, with a large number of cases occurring in low- and middle-income countries where access to early screening is limited. Colposcopy is a widely used method for examining the cervix and identifying abnormal tissue, but accurate interpretation often requires trained specialists. Our lab has developed an artificial intelligence-based framework to support cervical cancer screening by analyzing colposcopic images and assisting clinicians in identifying important diagnostic features.

The system analyzes medical images to detect key visual patterns related to cervical abnormalities and evaluates clinical indicators commonly used during screening. By learning from multiple datasets of colposcopic images and associated medical information, the framework can help assess cervical characteristics and support clinical decision-making. This approach aims to enhance screening accessibility, assist healthcare providers in diagnosis, and contribute to earlier detection of cervical abnormalities, particularly in resource-limited healthcare settings.

Lung cancer is the leading cause of cancer mortality worldwide, responsible for roughly one in five cancer deaths. In Bangladesh, it ranks among the most consequential cancers, placing third for new cases and second for cancer deaths in 2022. We focus on analyzing low-dose chest CT scans to automatically detect pulmonary nodules, including small or subtle findings that can be difficult to identify during routine review. By using 3D deep learning models that capture the spatial structure of lung nodules, our work aims to reduce radiologists' workload while improving the sensitivity of early cancer screening.

Beyond detection, we also study how AI can help determine whether a nodule is likely to be benign or malignant by learning from both its visual texture and its underlying CT intensity patterns. This allows us to move from simply finding nodules to providing more informative, risk-aware decision support for follow-up and diagnosis.