arXiv subjects

Amit Sethi

Publications and source records attributed to Amit Sethi.

At least 91 records · Page 5Linked to original sources

arXiv2023

Robust Semi-Supervised Learning for Histopathology Images through Self-Supervision Guided Out-of-Distribution Scoring

Semi-supervised learning (semi-SL) is a promising alternative to supervised learning for medical image analysis when obtaining good quality supervision for medical imaging is difficult. However, semi-SL assumes that the underlying distribution of unaudited data matches that of the few labeled samples, which is often violated in practical settings, particularly in medical images. The presence of out-of-distribution (OOD) samples in the unlabeled training pool of semi-SL is inevitable and can reduce the efficiency of the algorithm. Common preprocessing methods to filter out outlier samples may not be suitable for medical images that involve a wide range of anatomical structures and rare morphologies. In this paper, we propose a novel pipeline for addressing open-set supervised learning challenges in digital histology images. Our pipeline efficiently estimates an OOD score for each unlabelled data point based on self-supervised learning to calibrate the knowledge needed for a subsequent semi-SL framework. The outlier score derived from the OOD detector is used to modulate sample selection for the subsequent semi-SL stage, ensuring that samples conforming to the distribution of the few labeled samples are more frequently exposed to the subsequent semi-SL framework. Our framework is compatible with any semi-SL framework, and we base our experiments on the popular Mixmatch semi-SL framework. We conduct extensive studies on two digital pathology datasets, Kather colorectal histology dataset and a dataset derived from TCGA-BRCA whole slide images, and establish the effectiveness of our method by comparing with popular methods and frameworks in semi-SL algorithms through various experiments.

cs.CV↗

arXiv2023

EGFR Mutation Prediction of Lung Biopsy Images using Deep Learning

The standard diagnostic procedures for targeted therapies in lung cancer treatment involve histological subtyping and subsequent detection of key driver mutations, such as EGFR. Even though molecular profiling can uncover the driver mutation, the process is often expensive and time-consuming. Deep learning-oriented image analysis offers a more economical alternative for discovering driver mutations directly from whole slide images (WSIs). In this work, we used customized deep learning pipelines with weak supervision to identify the morphological correlates of EGFR mutation from hematoxylin and eosin-stained WSIs, in addition to detecting tumor and histologically subtyping it. We demonstrate the effectiveness of our pipeline by conducting rigorous experiments and ablation studies on two lung cancer datasets - TCGA and a private dataset from India. With our pipeline, we achieved an average area under the curve (AUC) of 0.964 for tumor detection, and 0.942 for histological subtyping between adenocarcinoma and squamous cell carcinoma on the TCGA dataset. For EGFR detection, we achieved an average AUC of 0.864 on the TCGA dataset and 0.783 on the dataset from India. Our key learning points include the following. Firstly, there is no particular advantage of using a feature extractor layers trained on histology, if one is going to fine-tune the feature extractor on the target dataset. Secondly, selecting patches with high cellularity, presumably capturing tumor regions, is not always helpful, as the sign of a disease class may be present in the tumor-adjacent stroma.

cs.CV↗

arXiv2023

Magnification Invariant Medical Image Analysis: A Comparison of Convolutional Networks, Vision Transformers, and Token Mixers

Convolution Neural Networks (CNNs) are widely used in medical image analysis, but their performance degrade when the magnification of testing images differ from the training images. The inability of CNNs to generalize across magnification scales can result in sub-optimal performance on external datasets. This study aims to evaluate the robustness of various deep learning architectures in the analysis of breast cancer histopathological images with varying magnification scales at training and testing stages. Here we explore and compare the performance of multiple deep learning architectures, including CNN-based ResNet and MobileNet, self-attention-based Vision Transformers and Swin Transformers, and token-mixing models, such as FNet, ConvMixer, MLP-Mixer, and WaveMix. The experiments are conducted using the BreakHis dataset, which contains breast cancer histopathological images at varying magnification levels. We show that performance of WaveMix is invariant to the magnification of training and testing data and can provide stable and good classification accuracy. These evaluations are critical in identifying deep learning architectures that can robustly handle changes in magnification scale, ensuring that scale changes across anatomical structures do not disturb the inference results.

eess.IV↗

arXiv2022

Artificial Intelligence-based Eosinophil Counting in Gastrointestinal Biopsies

Normally eosinophils are present in the gastrointestinal (GI) tract of healthy individuals. When the eosinophils increase beyond their usual amount in the GI tract, a patient gets varied symptoms. Clinicians find it difficult to diagnose this condition called eosinophilia. Early diagnosis can help in treating patients. Histopathology is the gold standard in the diagnosis for this condition. As this is an under-diagnosed condition, counting eosinophils in the GI tract biopsies is important. In this study, we trained and tested a deep neural network based on UNet to detect and count eosinophils in GI tract biopsies. We used connected component analysis to extract the eosinophils. We studied correlation of eosinophilic infiltration counted by AI with a manual count. GI tract biopsy slides were stained with H&E stain. Slides were scanned using a camera attached to a microscope and five high-power field images were taken per slide. Pearson correlation coefficient was 85% between the machine-detected and manual eosinophil counts on 300 held-out (test) images.

q-bio.QM↗

arXiv2022

Improving Mitosis Detection Via UNet-based Adversarial Domain Homogenizer

The effective localization of mitosis is a critical precursory task for deciding tumor prognosis and grade. Automated mitosis detection through deep learning-oriented image analysis often fails on unseen patient data due to inherent domain biases. This paper proposes a domain homogenizer for mitosis detection that attempts to alleviate domain differences in histology images via adversarial reconstruction of input images. The proposed homogenizer is based on a U-Net architecture and can effectively reduce domain differences commonly seen with histology imaging data. We demonstrate our domain homogenizer's effectiveness by observing the reduction in domain differences between the preprocessed images. Using this homogenizer, along with a subsequent retina-net object detector, we were able to outperform the baselines of the 2021 MIDOG challenge in terms of average precision of the detected mitotic figures.

cs.CV↗

arXiv2022

Deep Multi-Scale U-Net Architecture and Label-Noise Robust Training Strategies for Histopathological Image Segmentation

Although the U-Net architecture has been extensively used for segmentation of medical images, we address two of its shortcomings in this work. Firstly, the accuracy of vanilla U-Net degrades when the target regions for segmentation exhibit significant variations in shape and size. Even though the U-Net already possesses some capability to analyze features at various scales, we propose to explicitly add multi-scale feature maps in each convolutional module of the U-Net encoder to improve segmentation of histology images. Secondly, the accuracy of a U-Net model also suffers when the annotations for supervised learning are noisy or incomplete. This can happen due to the inherent difficulty for a human expert to identify and delineate all instances of specific pathology very precisely and accurately. We address this challenge by introducing auxiliary confidence maps that emphasize less on the boundaries of the given target regions. Further, we utilize the bootstrapping properties of the deep network to address the missing annotation problem intelligently. In our experiments on a private dataset of breast cancer lymph nodes, where the primary task was to segment germinal centres and sinus histiocytosis, we observed substantial improvement over a U-Net baseline based on the two proposed augmentations.

eess.IV↗

arXiv2022

Shallow Water Bathymetry Survey using an Autonomous Surface Vehicle

Accurate and cost effective mapping of water bodies has an enormous significance for environmental understanding and navigation. However, the quantity and quality of information we acquire from such environmental features is limited by various factors, including cost, time, security, and the capabilities of existing data collection techniques. Measurement of water depth is an important part of such mapping, particularly in shallow locations that could provide navigational risk or have important ecological functions. Erosion and deposition at these locations, for example, due to storms and erosion, can cause rapid changes that require repeated measurements. In this paper, we describe a low-cost, resilient, unmanned autonomous surface vehicle for bathymetry data collection using side-scan sonar. We discuss the adaptation of equipment and sensors for the collection of navigation, control, and bathymetry data and also give an overview of the vehicle setup. This autonomous surface vehicle has been used to collect bathymetry from the Powai Lake in Mumbai, India.

eess.SY↗

arXiv2022

Deriving Surface Resistivity from Polarimetric SAR Data Using Dual-Input UNet

Traditional survey methods for finding surface resistivity are time-consuming and labor intensive. Very few studies have focused on finding the resistivity/conductivity using remote sensing data and deep learning techniques. In this line of work, we assessed the correlation between surface resistivity and Synthetic Aperture Radar (SAR) by applying various deep learning methods and tested our hypothesis in the Coso Geothermal Area, USA. For detecting the resistivity, L-band full polarimetric SAR data acquired by UAVSAR were used, and MT (Magnetotellurics) inverted resistivity data of the area were used as the ground truth. We conducted experiments to compare various deep learning architectures and suggest the use of Dual Input UNet (DI-UNet) architecture. DI-UNet uses a deep learning architecture to predict the resistivity using full polarimetric SAR data by promising a quick survey addition to the traditional method. Our proposed approach accomplished improved outcomes for the mapping of MT resistivity from SAR data.

physics.geo-ph↗

arXiv2022

WaveMix: Resource-efficient Token Mixing for Images

Although certain vision transformer (ViT) and CNN architectures generalize well on vision tasks, it is often impractical to use them on green, edge, or desktop computing due to their computational requirements for training and even testing. We present WaveMix as an alternative neural architecture that uses a multi-scale 2D discrete wavelet transform (DWT) for spatial token mixing. Unlike ViTs, WaveMix neither unrolls the image nor requires self-attention of quadratic complexity. Additionally, DWT introduces another inductive bias -- besides convolutional filtering -- to utilize the 2D structure of an image to improve generalization. The multi-scale nature of the DWT also reduces the requirement for a deeper architecture compared to the CNNs, as the latter relies on pooling for partial spatial mixing. WaveMix models show generalization that is competitive with ViTs, CNNs, and token mixers on several datasets while requiring lower GPU RAM (training and testing), number of computations, and storage. WaveMix have achieved State-of-the-art (SOTA) results in EMNIST Byclass and EMNIST Balanced datasets.

cs.CV↗

arXiv2022

WSSAMNet: Weakly Supervised Semantic Attentive Medical Image Registration Network

We present WSSAMNet, a weakly supervised method for medical image registration. Ours is a two step method, with the first step being the computation of segmentation masks of the fixed and moving volumes. These masks are then used to attend to the input volume, which are then provided as inputs to a registration network in the second step. The registration network computes the deformation field to perform the alignment between the fixed and the moving volumes. We study the effectiveness of our technique on the BraTSReg challenge data against ANTs and VoxelMorph, where we demonstrate that our method performs competitively.

eess.IV↗

arXiv2022

Perceptual cGAN for MRI Super-resolution

Capturing high-resolution magnetic resonance (MR) images is a time consuming process, which makes it unsuitable for medical emergencies and pediatric patients. Low-resolution MR imaging, by contrast, is faster than its high-resolution counterpart, but it compromises on fine details necessary for a more precise diagnosis. Super-resolution (SR), when applied to low-resolution MR images, can help increase their utility by synthetically generating high-resolution images with little additional time. In this paper, we present a SR technique for MR images that is based on generative adversarial networks (GANs), which have proven to be quite useful in generating sharp-looking details in SR. We introduce a conditional GAN with perceptual loss, which is conditioned upon the input low-resolution image, which improves the performance for isotropic and anisotropic MRI super-resolution.

eess.IV↗

arXiv2021

Memory Efficient Adaptive Attention For Multiple Domain Learning

Training CNNs from scratch on new domains typically demands large numbers of labeled images and computations, which is not suitable for low-power hardware. One way to reduce these requirements is to modularize the CNN architecture and freeze the weights of the heavier modules, that is, the lower layers after pre-training. Recent studies have proposed alternative modular architectures and schemes that lead to a reduction in the number of trainable parameters needed to match the accuracy of fully fine-tuned CNNs on new domains. Our work suggests that a further reduction in the number of trainable parameters by an order of magnitude is possible. Furthermore, we propose that new modularization techniques for multiple domain learning should also be compared on other realistic metrics, such as the number of interconnections needed between the fixed and trainable modules, the number of training samples needed, the order of computations required and the robustness to partial mislabeling of the training data. On all of these criteria, the proposed architecture demonstrates advantages over or matches the current state-of-the-art.

cs.LG↗

arXiv2021

Vision Xformers: Efficient Attention for Image Classification

Although transformers have become the neural architectures of choice for natural language processing, they require orders of magnitude more training data, GPU memory, and computations in order to compete with convolutional neural networks for computer vision. The attention mechanism of transformers scales quadratically with the length of the input sequence, and unrolled images have long sequence lengths. Plus, transformers lack an inductive bias that is appropriate for images. We tested three modifications to vision transformer (ViT) architectures that address these shortcomings. Firstly, we alleviate the quadratic bottleneck by using linear attention mechanisms, called X-formers (such that, X in {Performer, Linformer, Nyströmformer}), thereby creating Vision X-formers (ViXs). This resulted in up to a seven times reduction in the GPU memory requirement. We also compared their performance with FNet and multi-layer perceptron mixers, which further reduced the GPU memory requirement. Secondly, we introduced an inductive bias for images by replacing the initial linear embedding layer by convolutional layers in ViX, which significantly increased classification accuracy without increasing the model size. Thirdly, we replaced the learnable 1D position embeddings in ViT with Rotary Position Embedding (RoPE), which increases the classification accuracy for the same model size. We believe that incorporating such changes can democratize transformers by making them accessible to those with limited data and computing resources.

cs.CV↗

arXiv2021

PAL : Pretext-based Active Learning

The goal of pool-based active learning is to judiciously select a fixed-sized subset of unlabeled samples from a pool to query an oracle for their labels, in order to maximize the accuracy of a supervised learner. However, the unsaid requirement that the oracle should always assign correct labels is unreasonable for most situations. We propose an active learning technique for deep neural networks that is more robust to mislabeling than the previously proposed techniques. Previous techniques rely on the task network itself to estimate the novelty of the unlabeled samples, but learning the task (generalization) and selecting samples (out-of-distribution detection) can be conflicting goals. We use a separate network to score the unlabeled samples for selection. The scoring network relies on self-supervision for modeling the distribution of the labeled samples to reduce the dependency on potentially noisy labels. To counter the paucity of data, we also deploy another head on the scoring network for regularization via multi-task learning and use an unusual self-balancing hybrid scoring function. Furthermore, we divide each query into sub-queries before labeling to ensure that the query has diverse samples. In addition to having a higher tolerance to mislabeling of samples by the oracle, the resultant technique also produces competitive accuracy in the absence of label noise. The technique also handles the introduction of new classes on-the-fly well by temporarily increasing the sampling rate of these classes.

cs.CV↗

arXiv2020

Fast, Self Supervised, Fully Convolutional Color Normalization of H&E Stained Images

Performance of deep learning algorithms decreases drastically if the data distributions of the training and testing sets are different. Due to variations in staining protocols, reagent brands, and habits of technicians, color variation in digital histopathology images is quite common. Color variation causes problems for the deployment of deep learning-based solutions for automatic diagnosis system in histopathology. Previously proposed color normalization methods consider a small patch as a reference for normalization, which creates artifacts on out-of-distribution source images. These methods are also slow as most of the computation is performed on CPUs instead of the GPUs. We propose a color normalization technique, which is fast during its self-supervised training as well as inference. Our method is based on a lightweight fully-convolutional neural network and can be easily attached to a deep learning-based pipeline as a pre-processing block. For classification and segmentation tasks on CAMELYON17 and MoNuSeg datasets respectively, the proposed method is faster and gives a greater increase in accuracy than the state of the art methods.

cs.CV↗

arXiv2020

Fast Design of Plasmonic Metasurfaces Enabled by Deep Learning

Metasurfaces is an emerging field that enables the manipulation of light by an ultra-thin structure composed of sub-wavelength antennae and fulfills an important requirement for miniaturized optical elements. Finding a new design for a metasurface or optimizing an existing design for a desired functionality is a computationally expensive and time consuming process as it is based on an iterative process of trial and error. We propose a deep learning (DL) architecture dubbed bidirectional autoencoder for nanophotonic metasurface design via a template search methodology. In contrast with the earlier approaches based on DL, our methodology addresses optimization in the space of multiple metasurface topologies instead of just one, in order to tackle the one to many mapping problem of inverse design. We demonstrate the creation of a Geometry and Parameter Space Library (GPSL) of metasurface designs with their corresponding optical response using our DL model. This GPSL acts as a universal design and response space for the optimization. As an example application, we use our methodology to design a multi-band gap-plasmon based half-wave plate metasurface. Through this example, we demonstrate the power of our technique in addressing the non-uniqueness problem of common inverse design. Our network converges aptly to multiple metasurface topologies for the desired optical response with a low mean absolute error between desired optical response and the optical response of topologies searched. Our proposed technique would enable fast and accurate design and optimization of various kinds of metasurfaces with different functionalities.

physics.optics↗

arXiv2020

Activation Functions: Do They Represent A Trade-Off Between Modular Nature of Neural Networks And Task Performance

Current research suggests that the key factors in designing neural network architectures involve choosing number of filters for every convolution layer, number of hidden neurons for every fully connected layer, dropout and pruning. The default activation function in most cases is the ReLU, as it has empirically shown faster training convergence. We explore whether ReLU is the best choice if one is aiming to desire better modularity structure within a neural network.

cs.LG↗

arXiv2020

Switching Loss for Generalized Nucleus Detection in Histopathology

The accuracy of deep learning methods for two foundational tasks in medical image analysis -- detection and segmentation -- can suffer from class imbalance. We propose a `switching loss' function that adaptively shifts the emphasis between foreground and background classes. While the existing loss functions to address this problem were motivated by the classification task, the switching loss is based on Dice loss, which is better suited for segmentation and detection. Furthermore, to get the most out the training samples, we adapt the loss with each mini-batch, unlike previous proposals that adapt once for the entire training set. A nucleus detector trained using the proposed loss function on a source dataset outperformed those trained using cross-entropy, Dice, or focal losses. Remarkably, without retraining on target datasets, our pre-trained nucleus detector also outperformed existing nucleus detectors that were trained on at least some of the images from the target datasets. To establish a broad utility of the proposed loss, we also confirmed that it led to more accurate ventricle segmentation in MRI as compared to the other loss functions. Our GPU-enabled pre-trained nucleus detection software is also ready to process whole slide images right out-of-the-box and is usably fast.

eess.IV↗