Sovanlal Mukherjee, Ph.D.

Location

Contact

SUMMARY

Sovanlal Mukherjee, Ph.D., studies artificial intelligence (AI) applications in medical imaging and radiology. With a background in electrical engineering and medical image processing, Dr. Mukherjee conducts research in radiomics-based machine learning (ML), deep learning-based segmentation and predictive modeling. He develops AI algorithms for early detection of pancreatic ductal adenocarcinoma and for volumetric pancreas and pancreatic tumor segmentation. He also uses multimodal imaging data for survival prediction. His prior research includes innovations in radiation therapy planning and image denoising for low-dose computerized tomography (CT) scans.

Focus areas

• Early pancreatic cancer detection. Dr. Mukherjee developed a radiomics-based ML model that detects early pancreatic ductal adenocarcinoma, a common form of pancreatic cancer, using prediagnostic CT scans. The model achieved 92.2% accuracy, significantly outperforming human radiologists. Tested on a large dataset, the model also demonstrated robustness against image perturbations, making it a promising tool for early diagnosis of pancreatic ductal adenocarcinoma.

  Building on this success, Dr. Mukherjee developed another radiomics-based ML model for the early detection of pancreatic ductal adenocarcinoma. This model, called REDMOD, was particularly trained on a high control-to-case ratio to reflect low-prevalence screening of pancreatic ductal adenocarcinoma. The model's performance has also significantly surpassed that of human radiologists.

• Volumetric pancreatic ductal adenocarcinoma segmentation using AI. To address challenges in pancreatic ductal adenocarcinoma and pancreatic tumor segmentation, Dr. Mukherjee created a bounding box-based convolutional neural network model that focuses on peritumoral anatomy for pancreatic ductal adenocarcinoma segmentation. This semiautomated approach achieved extremely high segmentation accuracy and generalized well across multi-institutional datasets, improving the reliability of imaging biomarkers for the disease.

  Based on the success of the semiautomated approach, Dr. Mukherjee developed another convolutional neural network model for fully automated pancreatic ductal adenocarcinoma segmentation. This automated version paves the way for various biomarker extraction from the pancreatic ductal adenocarcinoma region for potential applications, including survival analyses and prediction of treatment response and risk.

• Automated radiation therapy planning. Dr. Mukherjee designed a computationally efficient algorithm to integrate soft and hard dose-volume constraints into intensity-modulated radiation therapy fluence optimization. This two-phase approach enhances treatment planning by balancing clinical constraints and computational feasibility.
• Noise reduction in low-dose CT imaging. Dr. Mukherjee's comparative study of algorithms for noise reduction in cone-beam CT imaging — total variation with split Bregman and total variation with Nesterov — has demonstrated that the Nesterov method offers superior image quality and faster convergence. His work supports safer imaging practices by improving diagnostic clarity in low-dose scans.
• Thermo-acoustic tomography for cancer detection. Dr. Mukherjee conducted in silico studies of thermo-acoustic CT to reconstruct electrical conductivity profiles for cancer detection. Dr. Mukherjee uses finite element methods to demonstrate how microwave-induced acoustic signals can differentiate cancerous tissue from healthy tissue in both external and internal imaging geometries.

Significance to patient care

Dr. Mukherjee primarily develops AI tools that help medical teams find pancreatic cancer sooner when it is most treatable. These tools can spot signs of cancer in scans months before symptoms appear, giving patients a better chance at treatment. His research also creates faster and more accurate ways to plan radiation therapy and improve image quality in low-dose scans, making care safer and more effective.

Professional highlights

• Peer reviewer, Abdominal Radiology, 2022-present.
• Peer reviewer, Academic Radiology, 2022-present.
• Peer reviewer, European Journal of Radiology, 2022-present.
• Peer reviewer, Medical Physics, 2022-present.
• Peer reviewer, Scientific Reports, 2022-present.
• Peer reviewer, Pancreatology, 2022-present.
• Peer reviewer, Radiology, 2022-present.
• Mayo Clinic:
  • Radiology Research Grant Award, 2025.
  • Transformative Science Award, Pancreas Cancer Artificial Intelligence Team, 2025.
  • Strategic Pilot Study Grant Award, Mayo Clinic Comprehensive Cancer Center, 2024.
• Member, editorial board, Frontiers of Neuroscience, 2020.
• National Institutes of Health (NIH):
  • Travel award to attend Integrated Course in Biology and Physics of Radiation Oncology at Wayne State University, 2014.
  • Travel grant to attend NIH workshop, 2014.