• Our experimentation demonstrates that slide-level zero-shot transfer relying on textual ensemble learning outperforms randomly initialized MIL classification models. • We propose a novel framework termed MIL-Adapter that leverages histopathology vision-language models and enables robust slide-level classification in data- and parameter-efficient settings. • MIL-Adapter couples foundation model feature extraction and lightweight MIL models with vision-language adapters to perform consistent and interpretable prediction of the slides. • We conduct extensive validation of our framework in cancer subtyping tasks across a variety of organs and through diverse configurations of few-shot learning scenarios. Contrastive language-image pretraining has greatly enhanced visual representation learning and enabled zero-shot classification. Vision-language language models (VLM) have succeeded in few-shot learning by leveraging adaptation modules fine-tuned for specific downstream tasks. In computational pathology (CPath), accurate whole-slide image (WSI) prediction is crucial for aiding in cancer diagnosis, and multiple instance learning (MIL) remains essential for managing the gigapixel scale of WSIs. In the intersection between CPath and VLMs, the literature still lacks specific adapters that handle the particular complexity of the slides. To solve this gap, we introduce MIL-Adapter, a novel approach designed to obtain consistent slide-level classification under few-shot learning scenarios. In particular, our framework is the first to combine trainable MIL aggregation functions and lightweight visual-language adapters to improve the performance of histopathological VLMs. MIL-Adapter relies on textual ensemble learning to construct discriminative zero-shot prototypes. It is serves as a solid starting point, surpassing MIL models with randomly initialized classifiers in data-constrained settings. With our experimentation, we demonstrate the value of textual ensemble learning and the robust predictive performance of MIL-Adapter through diverse datasets and configurations of few-shot scenarios, while providing crucial insights on model interpretability. The code is publicly accessible in https://github.com/cvblab/MIL-Adapter .

Abstract Many current embedded systems comprise heterogeneous computing components including quite powerful GPUs, which enables their application across diverse sectors. This study demonstrates the efficient execution of a medium-sized self-supervised audio spectrogram transformer (SSAST) model on a low-power system-on-chip (SoC). Through comprehensive evaluation, including real time inference scenarios, we show that GPUs outperform multi-core CPUs in inference processes. Optimization techniques such as adjusting batch size, model compilation with TensorRT, and reducing data precision significantly enhance inference time, energy consumption, and memory usage. In particular, negligible accuracy degradation is observed, with post-training quantization to 8-bit integers showing less than 1% loss. This research underscores the feasibility of deploying transformer neural networks on low-power embedded devices, ensuring efficiency in time, energy, and memory, while maintaining the accuracy of the results.

Artificial intelligence (AI) agents encounter the problem of catastrophic forgetting when they are trained in sequentially with new data batches. This issue poses a barrier to the implementation of AI-based models in tasks that involve ongoing evolution, such as cancer prediction. Moreover, whole slide images (WSI) play a crucial role in cancer management, and their automated analysis has become increasingly popular in assisting pathologists during the diagnosis process. Incremental learning (IL) techniques aim to develop algorithms capable of retaining previously acquired information while also acquiring new insights to predict future data. Deep IL techniques need to address the challenges posed by the gigapixel scale of WSIs, which often necessitates the use of multiple instance learning (MIL) frameworks. In this paper, we introduce an IL algorithm tailored for analyzing WSIs within a MIL paradigm. The proposed Multiple Instance Class-Incremental Learning (MICIL) algorithm combines MIL with class-IL for the first time, allowing for the incremental prediction of multiple skin cancer subtypes from WSIs within a class-IL scenario. Our framework incorporates knowledge distillation and data rehearsal, along with a novel embedding-level distillation, aiming to preserve the latent space at the aggregated WSI level. Results demonstrate the algorithm's effectiveness in addressing the challenge of balancing IL-specific metrics, such as intransigence and forgetting, and solving the plasticity-stability dilemma.

Abstract Sound classification using neural networks has recently produced very accurate results. A large number of different applications use this type of sound classifiers such as controlling and monitoring the type of activity in a city or identifying different types of animals in natural environments. While traditional acoustic processing applications have been developed on high-performance computing platforms equipped with expensive multi-channel audio interfaces, the Internet of Things (IoT) paradigm requires the use of more flexible and energy-efficient systems. Although software-based platforms exist for implementing general-purpose neural networks, they are not optimized for sound classification, wasting energy and computational resources. In this work, we have used FPGAs to develop an ad hoc system where only the hardware needed for our application is synthesized, resulting in faster and more energy-efficient circuits. The results show that our developments are accelerated by a factor of 35 compared to a software-based implementation on a Raspberry Pi.

Cutaneous spindle cell (CSC) neoplasms are a group of tumors that represent a formidable diagnostic challenge for dermatopathologists. Digital pathology has enabled the application of new methods based on artificial intelligence to reduce the workload of pathologists’ daily practice. In this work, we propose a self-learning framework to detect tumor regions in histological images. The use of a teacher-model paradigm increases the annotated database while avoiding manual annotation. The pre-trained latent space of this model is then used in a second stage by another model to differentiate between leiomyomas (benign cases) and leiomyosarcomas (malignant cases). A contrastive learning approach allows separating the latent space of samples from different classes. This framework was tested on an independent database. This novel approach supposes a step forward in the CSC detection as the obtained results suggest (Acc = 0.90 and 0.8451, respectively).

Abstract Epigenetic alterations have an important role in the development of several types of cancer. Epigenetic studies generate a large amount of data, which makes it essential to develop novel models capable of dealing with large-scale data. In this work, we propose a deep embedded refined clustering method for breast cancer differentiation based on DNA methylation. In concrete, the deep learning system presented here uses the levels of CpG island methylation between 0 and 1. The proposed approach is composed of two main stages. The first stage consists in the dimensionality reduction of the methylation data based on an autoencoder. The second stage is a clustering algorithm based on the soft assignment of the latent space provided by the autoencoder. The whole method is optimized through a weighted loss function composed of two terms: reconstruction and classification terms. To the best of the authors’ knowledge, no previous studies have focused on the dimensionality reduction algorithms linked to classification trained end-to-end for DNA methylation analysis. The proposed method achieves an unsupervised clustering accuracy of 0.9927 and an error rate (%) of 0.73 on 137 breast tissue samples. After a second test of the deep-learning-based method using a different methylation database, an accuracy of 0.9343 and an error rate (%) of 6.57 on 45 breast tissue samples are obtained. Based on these results, the proposed algorithm outperforms other state-of-the-art methods evaluated under the same conditions for breast cancer classification based on DNA methylation data.

Nowadays, glaucoma is the leading cause of blindness worldwide. We propose in this paper two different deep-learning based approaches to address glaucoma detection just from raw circumpapillary OCT images. The first one is based on the development of convolutional neural networks (CNNs) trained from scratch. The second one lies in fine-tuning some of the most common state-of-the-art CNNs architectures. The experiments were performed on a private database composed of 93 glaucomatous and 156 normal B-scans around the optic nerve head of the retina, which were diagnosed by expert ophthalmologists. The validation results evidence that finetuned CNNs outperform the networks trained from scratch when small databases are addressed. Additionally, the VGG family of networks reports the most promising results, with an area under the ROC curve of 0.96 and an accuracy of 0.92, during the prediction of the independent test set.