Exploiting Patch Sizes and Resolutions for Multi-Scale Deep Learning in Mammogram Image Classification

Quintana, Gonzalo Iñaki; Li, Zhijin; Vancamberg, Laurence; Mougeot, Mathilde; Desolneux, Agnès; Muller, Serge

doi:10.3390/bioengineering10050534

Cited by 8 publications

(4 citation statements)

References 55 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Due to their tiny sizes (e.g., 100 × 100 pixels in 4000 × 3000 images), 3 keeping the high semantic details of the lesion is crucial to efficiently identify their morphology and associate a risk of malignancy. [16][17][18] On the contrary, degrading pixel information may lead to the creation of noise or bright artifacts that can be confused with MCs.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Annotation-free deep-learning framework for microcalcifications detection on mammograms

Terrassin,

Tardy,

Lauzeral

et al. 2024

Medical Imaging 2024: Computer-Aided Diagnosis

View full text Add to dashboard Cite

Breast cancer detection at an early stage significantly i ncreases t he chances o f r ecovery f or p atients. Mammography (MG) is one of the most popular non-invasive and high-resolution imaging allowing radiologists to depict early signs of the disease. Microcalcifications ( MCs) o ften o ccupy l ess t han 1 mm i n s ize a nd c an r epresent a high risk of suspicion depending on the spatial distribution, morphology, and their evolution over time. Their detection is challenging both the clinicians and computer-aided detection tools. In this work, we propose a novel annotation-free framework designed specifically f or t he M Cs d etection a nd t rained i n a self-supervised manner thanks to the generation of synthetic MCs. Inspired by the UNet3+ architecture, we reduced its number of parameters to make it applicable in practice and added multi-scale features to enrich fine-grained details with more global context information. Both multi-channel segmentation and multi-class classification t asks are implemented in a multi-scale output approach to catch MC of various sizes. We perform a comparison with several state-of-the-art methods, including different fl avors of Re sNet-22, Co nvNeXt, an d UN et3+. An analysis of classification a nd s egmentation p erformances h as b een d one, u sing t he G radient-weighted C lass Activation Mapping method to make classifiers v isually e xplainable. I n t his s tudy, w e u sed t wo p ublic d atasets, INBreast and Breast MicroCalcifications D ataset f or validation a nd t est p urposes. W e a chieved a n A UC s core o f 0 .93 in the characterization of malignant MCs while having a semantic segmentation precision of 0.70. To the best of our knowledge, we are the first s tudy c laiming s egmentation p erformances o n t he B MCD dataset.

show abstract

Section: Related Workmentioning

confidence: 99%

“…The authors conclude that a single size or resolution is not optimal for catching all lesions. 17 They combined patch classifiers of several sizes to generate the global image risk of malignancy prediction. All those methods focus on the image classification task.…”

Section: Related Workmentioning

confidence: 99%

Annotation-free deep-learning framework for microcalcifications detection on mammograms

Terrassin,

Tardy,

Lauzeral

et al. 2024

Medical Imaging 2024: Computer-Aided Diagnosis

View full text Add to dashboard Cite

show abstract

“…Tamanhos menores levaram à perda de informac ¸ões relevantes, devido à falta de contexto significativo, comprometendo assim a capacidade de capturar características distintivas da região e resultando em uma representac ¸ão inadequada. Por outro lado, tamanhos maiores resultaram em uma extrac ¸ão limitada de patches, comprometendo a capacidade do modelo em capturar detalhes finos presentes na região [Quintana et al 2023]. Durante a extrac ¸ão foram descartados patches com menos de 30% de pixels pertencentes a ROI, garantindo uma representac ¸ão mais robusta e informativa da região.…”

Section: Extrac ¸ãO De Patchesunclassified

Classificação da Camada Lipídica do Filme Lacrimal usando K-Means e Deep Learning

Dias Junior,

Cruz,

Diniz

2024

Anais Do XXIV Simpósio Brasileiro De Computação Aplicada À Saúde (SBCAS 2024)

View full text Add to dashboard Cite

A síndrome do olho seco, uma das doenças oftalmológicas mais comuns, apresenta desafios diagnósticos devido à sua natureza multifatorial. O diagnóstico tradicional, que envolve a classificação manual das imagens do filme lacrimal, enfrenta limitações devido à instabilidade deste filme. Este estudo propõe uma abordagem inovadora que combina aprendizagem não supervisionada para segmentação da região de interesse e deep learning para classificação de patches. Os resultados são promissores, com acurácia de 99,23% e F1-score de 99,16%, superando outras técnicas e estudos da literatura. Acredita-se que esta metodologia possa ser uma ferramenta valiosa para auxiliar os especialistas no diagnóstico da doença.

show abstract

“…TP denotes the number of correctly classified positive samples, TN denotes the number of correctly classified negative samples, FP denotes the number of samples incorrectly classified as positive, FN denotes the number of samples incorrectly classified as negative. Quintana et al (2023), a patch classifier is trained and then is extended to the whole image classifier. A multi-patch size and a multi-resolution approach are proposed to classify whole images as cancer and no cancer.…”

Section: Implementation Detailsmentioning

confidence: 99%

FSE-Net: feature selection and enhancement network for mammogram classification

Liao,

Wen,

et al. 2023

Phys. Med. Biol.

View full text Add to dashboard Cite

Objective. Early detection and diagnosis allow for intervention and treatment at an early stage of breast cancer. Despite recent advances in computer aided diagnosis systems based on convolutional neural networks for breast cancer diagnosis, improving the classification performance of mammograms remains a challenge due to the various sizes of breast lesions and difficult extraction of small lesion features. To obtain more accurate classification results, many studies choose to directly classify region of interest (ROI) annotations, but labeling ROIs is labor intensive. The purpose of this research is to design a novel network to automatically classify mammogram image as cancer and no cancer, aiming to mitigate or address the above challenges and help radiologists perform mammogram diagnosis more accurately. Approach. We propose a novel feature selection and enhancement network (FSE-Net) to fully exploit the features of mammogram images, which requires only mammogram images and image-level labels without any bounding boxes or masks. Specifically, to obtain more contextual information, an effective feature selection module is proposed to adaptively select the receptive fields and fuse features from receptive fields of different scales. Moreover, a feature enhancement module is designed to explore the correlation between feature maps of different resolutions and to enhance the representation capacity of low-resolution feature maps with high-resolution feature maps. Main results. The performance of the proposed network has been evaluated on the CBIS-DDSM dataset and INbreast dataset. It achieves an accuracy of 0.806 with an AUC of 0.866 on the CBIS-DDSM dataset and an accuracy of 0.956 with an AUC of 0.974 on the INbreast dataset. Significance. Through extensive experiments and saliency map visualization analysis, the proposed network achieves the satisfactory performance in the mammogram classification task, and can roughly locate suspicious regions to assist in the final prediction of the entire images.

show abstract

Exploiting Patch Sizes and Resolutions for Multi-Scale Deep Learning in Mammogram Image Classification

Cited by 8 publications

References 55 publications

Annotation-free deep-learning framework for microcalcifications detection on mammograms

Annotation-free deep-learning framework for microcalcifications detection on mammograms

Classificação da Camada Lipídica do Filme Lacrimal usando K-Means e Deep Learning

FSE-Net: feature selection and enhancement network for mammogram classification

Contact Info

Product

Resources

About