I am passionate about and focused on bringing AI-based healthcare solutions that make a difference. My work spans a broad portfolio of AI applications in Medical Imaging that aid in fast and efficient clinical decision-making. 

AI-Rad Companion Chest X-ray

Global Product Manager, Siemens Healthineers, Germany

The application helps to detect radiographic findings on Chest X-ray images with the use of AI. The algorithm can characterize and highlight the findings of pulmonary lesions, pleural effusion, pneumothorax, consolidation, and atelectasis. The output from the AI-Rad Companion Chest X-ray is used in concurrent-read mode to support radiologists in their differential diagnosis and clinical decision-making.

Screenshot 2021-05-15 at 13.08.42.png

Computational Neuroimaging

Research Scientist,

German Center for Neurodegenerative Diseases, Bonn, Germany

  • Contribution

    • Fully convolutional architectures for training and quality control with limited annotations through intelligent transfer learning and Bayesian Deep Learning.

    • Learning based image reconstruction from sparsely sampled complex valued k-space MRI data through Complex Fully Convolutional Networks.

    • Evaluated robustness of encoder-decoder architectures to adversarial attacks.

  • Proof-of-concept: Robust and fast whole brain segmentation and cortical parcellation (102 semantic classes) for investigating normal brain development, aging and neuro-degeneration.

  • Impact: Achieved whole brain segmentation in under 20 seconds. FreeSurfer neuro-image analysis pipeline accelerated ten-fold.  


Evidence Based Image Understanding

Doctoral Candidate,
CAMPAR (Prof. Dr. Nassir Navab)

Technische Universität München 

  • Contribution: Developed fast and scalable learning to hash methods (binary semantic indexing) to explore and exploit large-scale heterogeneous medical databases. 

  • Proof-of-concept: Demonstrated for semantics-preserving retrieval of neuronal images, searching chest X-rays with co-morbidities and multiple instance retrieval in mammography and histopathology.

  • Impact: Effectively overcame prohibitive computational complexity and memory overheads associated with image retrieval.   



Retinal Layer Segmentation

Graduate Student, CAMPAR (Prof. Dr. Nassir Navab)

Technische Universität München 

Fully convolutional deep architecture, termed ReLayNet, for end-to-end segmentation of retinal layers and fluid masses in eye OCT scans.


Computational Cardiology

Graduate Student, CAMPAR (Prof. Dr. Nassir Navab)

Technische Universität München 

  • Contribution: Investigated automated techniques for tissue-characterization and segmentation (with domain adaptation) through computational modeling of modality physics (tissue-energy interaction).

  • Proof-of-concept: Demonstrated for in vivo tissue characterization in intravascular ultrasound and optical coherence tomography for assessment of atherosclerotic plaques.   

  • Impact: Effective in learning under high anatomical variability, extreme class imbalance and presence of pathology associated with in vivo tissue characterization.

IVUS-Domain Adaptation.jpg