Sep 2023 — Mar 2024
Graph Convolutional Network based Multi-modal Learning Approaches for Breast Cancer Survival Prediction
Developed advanced multi-modal deep learning frameworks to address the critical challenge of accurately
predicting breast cancer prognosis and survival outcomes.
The first framework, MGCN-CalRF, utilizes multimodal Graph Convolutional Networks to extract modality-specific relational features and
integrates them using a calibrated Random Forest classifier for prognosis prediction. The second framework, CAGCL, applies graph contrastive
learning to refine feature embeddings and incorporates a cross-modality attention mechanism to model complementary information across heterogeneous
data modalities for survival prediction.
IIT Patna
Collaborators: Dr. Susmita Palmal, Prof. Somanath Tripathy, Dr. Sriparna Saha
Sep 2023 — Jan 2024
Multi-modal Breast-Cancer Survival Estimation and Multi-modal Skin Lesion classification.
Literature Survey on deep multi-modal architecture for accurately estimating the survival of breast cancer patients.
The architecture is aimed to utilize six different modalities from multi-omics, clinical, and histopathology profiles of TCGA-BRCA data.
We also worked on multi-modal deep learning-based architectures for skin lesion classification using dermoscopy
images and demographic information of the patients.
KMITL Thailand, BITS Pilani, IIT Patna
Collaborators: Dr. Sriparna Saha, Prof. Kitsuchart Pasupa, Prof. Snehanshu Saha, Dr. Archana Mathur
Jul 2022 — Jan 2023
Towards Stability and Robustness of AI-based Multi-modal Machine Learning Classifiers for Breast Cancer Prognosis Prediction
Developed some state-of-the-art machine learning classifiers by incorporating additional sources of
information from six different modalities which include clinical, genetic, and histopathological tissue images
for breast cancer survival prediction. Improving stability and robustness of these classifiers by introducing
some deep learning-based feature extraction techniques followed by a proposal of advanced kernel function
for the Support Vector Machine classifier
BITS Pilani, IIT Patna
Collaborators: Prof. Snehanshu Saha, Dr. Sriparna Saha, Dr. Archana Mathur
Jan 2022 — May 2023
Developing Ensemble techniques for Breast Cancer Survival Prognosis
Proposed variors ensemble architectures for the multi-modal breast cancer surviavl prognosis. "DeSuFEn"
is one of them which uses rewarding algorithm to counter the effect of class imbalance and fuzzy ensemble to dynamically allocate
the importance of each base classifier towards getting better prediction accuracies. Further, enhancement
of fuzzy ensemble with the help of two non-linear functions acting as deviation and support for breast
cancer survival prediction.
IIT Patna
Collaborators: Dr. Sriparna Saha
Jan 2021 — Dec 2021
AI-based Generative Models to Handle Missing Data in Multi-modal Framework for Breast Cancer Prognosis.
Developed a deep learning-based generative model for handling cases where patients are not having
information from all the modalities. The generative model uses the power of attention technique and
Generative Adversarial Networks to generate the missing data from the available data. Further developed
deep learning-based classifiers using the generated and available data towards breast cancer survival
prediction.
IIT Patna
Collaborators: Dr. Sriparna Saha
Jul 2018 — Dec 2020
Deep Learning Multi-modal Architectures for Breast Cancer Prognosis
Analysed different modalities of information from gene expression, copy number variation, and clinical
details of breast cancer patients towards the prognosis prediction task. Developed various deep learning
architectures that incorporate the concept of convolutional neural networks, attention mechanisms, and
stacking of machine learning classifiers over deep learning architectures to classify the patients as long-term
vs short-term survivors.
IIT Patna
Collaborators: Dr. Sriparna Saha