A5012268461- AI in cancer detection
- Radiomics and Machine Learning in Medical Imaging
- Digital Imaging for Blood Diseases
- Cervical Cancer and HPV Research
- Medical Image Segmentation Techniques
- Medical Imaging and Analysis
- Cell Image Analysis Techniques
- Breast Cancer Treatment Studies
- Brain Tumor Detection and Classification
- COVID-19 diagnosis using AI
- Machine Learning and Data Classification
- Advanced X-ray and CT Imaging
- Colorectal Cancer Screening and Detection
- Advanced Adaptive Filtering Techniques
- Advanced Neural Network Applications
- Lung Cancer Diagnosis and Treatment
- Machine Fault Diagnosis Techniques
- Remote Sensing and LiDAR Applications
- Speech and Audio Processing
- Genital Health and Disease
- Automated Road and Building Extraction
- Infrared Thermography in Medicine
- Machine Learning in Bioinformatics
- HER2/EGFR in Cancer Research
- Air Quality Monitoring and Forecasting
Indian Institute of Technology Bombay
2019-2024
The University of Adelaide
2024
Indian Institute of Technology Gandhinagar
2016
Detecting various types of cells in and around the tumor matrix holds a special significance characterizing micro-environment for cancer prognostication research. Automating tasks detecting, segmenting, classifying nuclei can free up pathologists' time higher value reduce errors due to fatigue subjectivity. To encourage computer vision research community develop test algorithms these tasks, we prepared large diverse dataset nucleus boundary annotations class labels. The has over 46,000 from...
Several therapeutically important mutations in cancers are economically detected using immunohistochemistry (IHC), which highlights the overexpression of specific antigens associated with mutation. However, IHC panels can be imprecise and relatively expensive low-income settings. On other hand, although hematoxylin eosin (H&E) staining used to visualize general tissue morphology is a routine low cost, it does not highlight any antigen or mutation.Using human epidermal growth factor receptor...
The suggestion that the systemic immune response in lymph nodes (LNs) conveys prognostic value for triple-negative breast cancer (TNBC) patients has not previously been investigated large cohorts. We used a deep learning (DL) framework to quantify morphological features haematoxylin and eosin-stained LNs on digitised whole slide images. From 345 patients, 5,228 axillary LNs, cancer-free involved, were assessed. Generalisable multiscale DL frameworks developed capture germinal centres (GCs)...
<p>Supplementary Figure S2. Paired scatterplots comparing 17 Luminal A cases from the test set obtained hospitals that did not contribute for model training versus 213 remaining in set. Correlations between iLumA% image and A. pLumA representing transcriptome purity, B. Oncotype DX score, C. Mammaprint D. Expression of GRB7, which plays an important role HER2 aggressiveness.</p>
<p>Schematic overview of the method for quantifying subtype heterogeneity in held-out whole-slide H&E images from TCGA-BRCA cohort. Conv-layer, convolutional layer; SSGCE, sample-specific generalized cross-entropy.</p>
<p>Supplementary Figure S1. Scatterplot of LumA proportion by transcriptomic analysis versus percentage tumor image patches classified as the DNN model - including held-out cases with nonLumA PAM50 assignment (n = 256). Best-fitting regression line shown, 95% confidence band. Horizontal lines depict quartile thresholds for number cases.</p>
<p>Clinical and molecular features of PAM50 LumA breast cancers in the test set according to quartile tumor area classified as by deep learning model (total <i>n</i> = 230)</p>
<div>Abstract<p>Intratumor heterogeneity (ITH) presents challenges for precision oncology, but methods its spatial quantification, scalable at population levels, do not exist. Based on previous work showing that the admixture of PAM50 subtype can be measured from bulk tissue using transcriptomic data, we trained a deep convolutional neural network to quantify ITH in luminal A (LumA) breast cancer routinely stained whole-slide images. We tested hypothesis detected images was...
<p>PFS for the test cohort of PAM50-assigned LumA breast cancer, comparing pure vs. admixed cases based on proportion tumor image classified as by deep learning model</p>
<p>Diagram depicting the allocation of breast cancer WSIs for analysis. Phase A: filtering TCGA images. B: 680 cases were divided by assigned subtype, and then subsets with transcriptomically pure subtype adherence randomly split into training, validation, test subgroups. C: Pure each held out initial testing. The final set used all LumA that out, representing full range transcriptomic purity. IDC, invasive ductal carcinoma; ILC, lobular carcinoma.</p>
<p>Supplementary Figure S2. Paired scatterplots comparing 17 Luminal A cases from the test set obtained hospitals that did not contribute for model training versus 213 remaining in set. Correlations between iLumA% image and A. pLumA representing transcriptome purity, B. Oncotype DX score, C. Mammaprint D. Expression of GRB7, which plays an important role HER2 aggressiveness.</p>
<p>Supplementary Figure S1. Scatterplot of LumA proportion by transcriptomic analysis versus percentage tumor image patches classified as the DNN model - including held-out cases with nonLumA PAM50 assignment (n = 256). Best-fitting regression line shown, 95% confidence band. Horizontal lines depict quartile thresholds for number cases.</p>
Histopathology whole slide images (WSIs) are being widely used to develop deep learning-based diagnostic solutions, especially for precision oncology. Most of these softwares vulnerable biases and impurities in the training test data which can lead inaccurate diagnoses. For instance, WSIs contain multiple types tissue regions, at least some might not be relevant diagnosis. We introduce HistoROI, a robust yet lightweight classifier segregate WSI into six broad regions -- epithelium, stroma,...
Abstract Deep learning (DL)‐based interpretation of medical images has reached a critical juncture expanding outside research projects into translational ones, and is ready to make its way the clinics. Advances over last decade in data availability, DL techniques, as well computing capabilities have accelerated this journey. Through journey, today we better understanding challenges pitfalls wider adoption clinical care, which, according us, should will drive advances field next few years....
Our goal is to develop technique for assessing bathymetry maps, which show the topography of floors water-bodies, using satellite or aerial imagery. The advent deep neural networks has enabled use new techniques in analysing and creating depth maps from high resolution images. In this paper we report a pilot study exploring potential learning architecture by framing problem as pixel-wise classification task. We took Sentinel-2 bands image independent measurements Humminbird™data. data...
The accuracy of deep learning classifiers trained using the cross entropy loss function suffers even when a fraction training labels are wrong or input images uninformative. Training and for computational pathology often noisy due to difficulty in signal localization certain disease classifications being subjective discretization underlying continuums conditions. For robust label noise, we propose modified sample-specific version generalized loss. We take advantage bootstrapping properties...
Performance of deep learning algorithms decreases drastically if the data distributions training and testing sets are different. Due to variations in staining protocols, reagent brands, habits technicians, color variation digital histopathology images is quite common. Color causes problems for deployment learning-based solutions automatic diagnosis system histopathology. Previously proposed normalization methods consider a small patch as reference normalization, which creates artifacts on...