Improving Multiple-Instance Learning via Disambiguation by Considering Generalization

Multiple instance learning (MIL) is a new paradigm in machine learning that deals with classification of bags of instances, as opposed to the traditional view that aims at learning from single instances. In a typical MIL setting, a negative bag is composed of only negative instances. On the other hand, a bag is considered positive if it contains at least one positive instance. This learning approach provides a natural way of modeling several pattern recognition and computer vision problems, e.g. protein, document and image classification and object recognition, which inherently require learning under ambiguity. In many cases, MIL approaches performs better than the standard single instance learning (SIL) methods. One of the general approaches to MIL is to transform a given MIL problem into a corresponding SIL problem. This transformation is mainly done with selecting a set of representative instances from the training bags, and these group of studies basically differ from each other on how they perform this instance selection step. In this study, we revisit such a MIL approach called MILIS with a different instance selection mechanism. The experimental results show that the proposed approach performs better on MIL benchmark data sets as compared to the original algorithm.

Multiple instance learning: A survey of problem characteristics and applications

We address the task of automatic detection of lesions caused by multiple myeloma (MM) in femurs or other long bones from CT data. Such detection is already an important part of the multiple myeloma diagnosis and staging. However, it is so far performed mostly manually, which is very time consuming. We formulate the detection as a multiple instance learning (MIL) problem, where instances are grouped into bags and only bag labels are available. In our case, instances are regions in the image and bags correspond to images. This has the advantage of requiring only subject-level annotation (ground truth), which is much easier to get than voxel-level manual segmentation. We consider a generalization of the standard MIL formulation where we introduce a threshold on the number of required positive instances in positive bags. This corresponds better to the classification procedure used by the radiology experts and is more robust with respect to false positive instances. We extend several existing MIL algorithms to solve the generalized case by estimating the threshold during learning. We compare the proposed methods with the baseline method on a dataset of 220 subjects. We show that the generalized MIL formulation outperforms standard MIL methods for this task. For the task of distinguishing between healthy controls and MM patients with infiltrations, our best method makes almost no mistakes with a mean AUC of 0.982 and F1 = 0.965. We outperform the baseline method significantly in all conducted experiments.

Robust multiple-instance learning ensembles using random subspace instance selection

Multiple-Instance Learning (MIL) represents a new class of supervised learning tasks, where training examples are bags of instances with labels only available for the bags. To solve the instance label ambiguity, instance selection based MIL models were proposed to convert bag learning to traditional vector learning. However, existing MIL instance selection approaches are all based on the instances inside the bags. In this case, at the original instance space, those potential informative instances, which do not occur in the bags are discarded. In this paper, we propose a novel learning method, MILEIS (Multiple-Instance Learning with Evolutionary Instance Selection), to adaptively determine the informative instances for feature mapping. The unique evolutionary search mechanism, including instance initialization, mutation, and crossover, ensures that MILEIS can adjust itself to the data without explicit specification of functional or distributional form for the underlying model. By doing so, MILEIS can also take full advantage of those creative informative instances to help feature mapping in an accurate way. Experiments and comparisons on real-world applications demonstrate the effectiveness of the proposed method.

We address the task of detecting cancer in histological slide images based on training with weak, slide- and patch-level annotations, which are considerably easier to obtain than pixel-level annotations. we use CNN based patch-level descriptors and formulate the image classification task as a generalized multiple instance learning (MIL) problem. The generalization consists of requiring a certain number of positive instances in positive bags, instead of just one as in standard MIL. The descriptors are learned on a small number of patch-level annotations, while the MIL layer uses only image-level patches for training.

Fuzzy logic is a powerful tool to model knowledge uncertainty, measurements imprecision, and vagueness. However, there is another type of vagueness that arises when data have multiple forms of expression. This is the case for multiple instance learning problems (MIL). In MIL, an object is represented by a collection of instances, called a bag. Labels of bags are known but not those of individual instances. In previous work, we have introduced MI-ANFIS, a Multiple Instance-ANFIS, that extends the standard Adaptive Neuro Fuzzy Inference System (ANFIS) to handle reasoning with bags of instances as input and capable of learning from ambiguously labeled data. In this paper, first, we present a regularization technique, we called Rule Dropout, and show that it could be used to train MI-ANFIS systems with better generalization. Next, we develop a multi class MI-ANFIS (MCMI-ANFIS), that could be used to solve multiple class classification problems effectively. The proposed MCMI-ANFIS is tested and validated using a benchmark data set suitable for multiple class MIL problems.

Probabilistic smooth attention for deep multiple instance learning in medical imaging

CIMIL-CRC: A clinically-informed multiple instance learning framework for patient-level colorectal cancer molecular subtypes classification from H&E stained images.

Multiple instance learning (MIL) has emerged as the best solution for whole slide image (WSI) classification. It consists of dividing each slide into patches, which are treated as a bag of instances labeled with a global label. MIL includes two main approaches: instance-based and embedding-based. In the former, each patch is classified independently, and then, the patch scores are aggregated to predict the bag label. In the latter, bag classification is performed after aggregating patch embeddings. Even if instance-based methods are naturally more interpretable, embedding-based MILs have usually been preferred in the past due to their robustness to poor feature extractors. Recently, the quality of feature embeddings has drastically increased using self-supervised learning (SSL). Nevertheless, many authors continue to endorse the superiority of embedding-based MIL. We conduct 710 experiments across 4 datasets, comparing 10 MIL strategies, 6 self-supervised methods with 4 backbones, 4 foundation models, and various pathology-adapted techniques. Furthermore, we introduce 4 instance-based MIL methods, never used before in the pathology domain. We show that with a good SSL feature extractor, simple instance-based MILs, with very few parameters, obtain similar or better performance than complex, state-of-the-art (SOTA) embedding-based MIL methods, setting new SOTA results on the BRACS and Camelyon16 datasets. As simple instance-based MIL methods are naturally more interpretable and explainable to clinicians, our results suggest that more effort should be put into well-adapted SSL methods for WSI rather than into complex embedding-based MIL methods.

Due to the trade-off between spatial resolution of the imagery and satellite revisit times, the research topics using multiresolution remote sensing data have attracted much attention. In recent years, the techniques for improving the visibility of multi-resolution imagery have been proposed, including pan-sharpening and super-resolution. However, there are relatively few studies on the techniques of improving the model performance on low-resolution imagery by referring to detailed information from the high-resolution imagery during training time. To tackle this type of task, domain adaptation has been proposed in the field of computer vision to adapt a model trained on one dataset to another with different properties. Yet, domain adaptation for multi-resolution data is difficult due to the scale variation in addition to differences from the camera sensors. In this study, we propose a new approach for multi-resolution modeling that combines the major techniques, semi-supervised domain adaptation (SSDA) and multiple instance learning (MIL). Under the MIL framework, a large scene can be regarded as a bag of instances (e.g., image patches), and information from different receptive field sizes can be exploited. We conducted experiments on a dataset of Japanese oak wilt, which is known to have severe forest damage with two different optical satellite imagery, SPOT6&7 satellite imageries (1.5m) and Pleiades-1B satellite imagery (0.5m). The proposed method improves the discrimination accuracy of the low-resolution model compared to the standard SSDA technique. This obtained result reveals the potential usefulness of MIL for effective multi-resolution modeling.

Since real-world data sets usually contain large instances, it is meaningful to develop efficient and effective multiple instance learning (MIL) algorithm. As a learning paradigm, MIL is different from traditional supervised learning that handles the classification of bags comprising unlabeled instances. In this paper, a novel efficient method based on extreme learning machine (ELM) is proposed to address MIL problem. First, the most qualified instance is selected in each bag through a single hidden layer feedforward network (SLFN) whose input and output weights are both initialed randomly, and the single selected instance is used to represent every bag. Second, the modified ELM model is trained by using the selected instances to update the output weights. Experiments on several benchmark data sets and multiple instance regression data sets show that the ELM-MIL achieves good performance; moreover, it runs several times or even hundreds of times faster than other similar MIL algorithms.

Histopathology image analysis plays an important role in the treatment and diagnosis of cancer. However, analysis of whole slide images (WSI) with deep learning is challenging given that the duration of pixel-level annotations is laborious and time consuming. To address this, recent methods have considered WSI classification as a Multiple Instance Learning (MIL) problem often with a multi-stage process for learning instance and slide level features. Currently, most methods focus on either instance-selection or instance prediction-aggregation that often fails to generalize and ignores instance relations. In this work, we propose a MIL-based method to jointly learn both instance- and bag-level embeddings in a single framework. In addition, we propose a center loss that maps embeddings of instances from the same bag to a single centroid and reduces intra-class variations. Consequently, our model can accurately predict instance labels and leverages robust hierarchical pooling of features to obtain bag-level features without sacrificing accuracy. Experimental results on curated colon datasets show the effectiveness of the proposed methods against recent state-of-the-art methods.

Horror scene detection is a research problem that has much practical use. The supervised method requires the training data to be labeled manually, which can be tedious and onerous. In this paper, a more challenging setting of the problems without complete information on data labels is investigated. In particular, as the horror scene is characterized by multiple features, this problem is formulated as a special multiple instance learning (MIL) problem - Multiple Grouped Instance Learning (MGIL), which requires partial labeled training. To solve the MGIL problem, a learning method is proposed - Multiple Distance-Expectation Maximization Diversity Density (MD-EMDD). Additionally, a survey is conducted to collect people's opinions based on the definition of horror scenes. Combined with the survey results, Labeled with Ranking - MD - EMDD is proposed and demonstrated better results when compared to the traditional MIL algorithm and close to performance achieved by supervised method.

The computer aided diagnosis (CAD) problems of detecting potentially diseased structures from medical images are typically distinguished by the following challenging characteristics: extremely unbalanced data between negative and positive classes; stringent real-time requirement of online execution; multiple positive candidates generated for the same malignant structure that are highly correlated and spatially close to each other. To address all these problems, we propose a novel learning formulation to combine cascade classification and multiple instance learning (MIL) in a unified min-max framework, leading to a joint optimization problem which can be converted to a tractable quadratically constrained quadratic program and efficiently solved by block-coordinate optimization algorithms. We apply the proposed approach to the CAD problems of detecting pulmonary embolism and colon cancer from computed tomography images. Experimental results show that our approach significantly reduces the computational cost while yielding comparable detection accuracy to the current state-of-the-art MIL or cascaded classifiers. Although not specifically designed for balanced MIL problems, the proposed method achieves superior performance on balanced MIL benchmark data such as MUSK and image data sets.