Medical Image Analysis
Representation learning for clinically meaningful classification and segmentation across CT, MRI, and radiographic imaging.
MSc Computer Science · University of Victoria
Nidita Roy
Medical image analysis · class-incremental learning · multimodal healthcare AI
I study how medical-imaging systems can acquire new clinical tasks while retaining useful prior knowledge. My current thesis direction focuses on class-incremental learning with modality-aware representation primitives for CT and MRI, while building a research agenda around retention, forward transfer, and reliable learning from heterogeneous clinical data.
FIG. 01Nidita Roy · MSc researcher
medical imagingrepresentation learningsequential adaptation
8.50/9.0Cumulative MSc GPA
MICAD 2026Accepted continual-learning paper
125 hFall 2026 TA appointment
FGS × 22025–26 and 2026–27 fellowship recipient
Affiliations & profiles
Research focus
Class-Incremental Learning
Learning sequential clinical tasks while measuring retention, forward transfer, and task-level performance.
Modality-Aware Representations
Task-conditioned representation primitives that account for CT intensity structure and MRI pulse-sequence variation.
Multimodal Healthcare AI
Combining images with demographic, physiological, and structured clinical information.
Current research
Preliminary MSc thesis direction
In progress
Class-Incremental Learning with Modality-Aware Representation Primitives
The thesis investigates whether modality-aware representations can make continual medical-imaging systems more robust across sequential tasks. The planned work develops task-conditioned CT representations, MRI pulse-sequence-aware representations, and continual-learning evaluations centered on retention, forward transfer, and task performance.
Selected research outputs
What Limits Continual Learning for Abdominal-Trauma CT Detection?
Co-lead and corresponding-author diagnostic study of forgetting versus representation and label bottlenecks.
Paper code →Spinal Disease Detection via Deep Multimodal Fusion
Lead-author multimodal study combining X-ray imagery with demographic features for spinal-disease classification.
IEEE Xplore →Patient survival prediction using actigraphy and clinical data
Multimodal deep learning using wearable activity and clinical information for survival-outcome prediction.
Open article →Education
2025–present
MSc in Computer Science, University of Victoria
Research in continual medical imaging under the supervision of Dr. Ashery Mbilinyi. Cumulative GPA: 8.50/9.0.
2025–2027
Faculty of Graduate Studies Fellowship
FGS fellowship recipient for 2025–2026 and 2026–2027; the 2026–2027 award is CAD $20,000.
Graduate preparation
CSC 502 Systems for Massive Datasets (97% A+); CSC 503 Data Mining (93% A+); CSC 581B Deep Learning for Medical Image Analysis (92% A+); CSC 581C Online Learning (81% A−).
PhD opportunities and research collaboration
I am interested in research on reliable and adaptive medical AI, especially continual learning, medical computer vision, multimodal learning, and learning from heterogeneous clinical data.
Get in touchEmail (institutional email)Email (personal email)