A real use case of semi-supervised learning for mammogram classification in a local clinic of Costa Rica

Abstract

The implementation of deep learning-based computer-aided diagnosis systems for the classification of mammogram images can help in improving the accuracy, reliability, and cost of diagnosing patients. However, training a deep learning model requires a considerable amount of labelled images, which can be expensive to obtain as time and effort from clinical practitioners are required. To address this, a number of publicly available datasets have been built with data from different hospitals and clinics, which can be used to pre-train the model. However, using models trained on these datasets for later transfer learning and model fine-tuning with images sampled from a different hospital or clinic might result in lower performance. This is due to the distribution mismatch of the datasets, which include different patient populations and image acquisition protocols. In this work, a real-world scenario is evaluated where a novel target dataset sampled from a private Costa Rican clinic is used, with few labels and heavily imbalanced data. The use of two popular and publicly available datasets (INbreast and CBIS-DDSM) as source data, to train and test the models on the novel target dataset, is evaluated. A common approach to further improve the model's performance under such small labelled target dataset setting is data augmentation. However, often cheaper unlabelled data is available from the target clinic. Therefore, semi-supervised deep learning, which leverages both labelled and unlabelled data, can be used in such conditions. In this work, we evaluate the semi-supervised deep learning approach known as MixMatch, to take advantage of unlabelled data from the target dataset, for whole mammogram image classification. We compare the usage of semi-supervised learning on its own, and combined with transfer learning (from a source mammogram dataset) with data augmentation, as also against regular supervised learning with transfer learning and data augmentation from source datasets. It is shown that the use of a semi-supervised deep learning combined with transfer learning and data augmentation can provide a meaningful advantage when using scarce labelled observations. Also, we found a strong influence of the source dataset, which suggests a more data-centric approach needed to tackle the challenge of scarcely labelled data. We used several different metrics to assess the performance gain of using semi-supervised learning, when dealing with very imbalanced test datasets (such as the G-mean and the F2-score), as mammogram datasets are often very imbalanced.

A Systematic Literature Review on Mammography: Deep Learning in Redefining Breast Cancer Diagnosis for the Asian Perspective

“…In addition, new techniques have emerged in the latest studies, such as attention based CNNs which use the attention block to assist the network in focusing on discriminating the characteristics of images (Al-Hejri et al, 2022;Songsaeng et al, 2021;Wang et al, 2022;Sun et al, 2023;Hina et al, 2020;Zhang et al, 2020), Generative Adversarial Networks Deep Adversarial Domain adaptation (Wang et al, 2021a) which uses domain adaptation for classification, Generative Adversarial Network in which the generator generates realistic images and the discriminator differentiates between real and synthetic data, along with performing classification tasks (Park et al, 2023;Ponraj and Canessane, 2023b;Shivhare and Saxena, 2022), and Student teacher InterNRL (Wang et al, 2023) in which the student serves as a prototype-based classifier and teacher represents a image classifier. Auto Encoders (Hamza, 2023) trained on labeled data and contributed to unsupervised feature learning and supervised classification, while Visual Transformers (ViT) (Chen et al, 2022;Xia et al, 2023;Prodan et al, 2023) utilize a transformer-style architecture across patches of the image, and semi-supervised (Calderon-Ramirez et al, 2022) were also used to overcome the missing annotation problem. We will examine several other kinds of DL models in the Outcomes sections that follow.…”

“…Wavelet Neural network (WNN) (Calderon-Ramirez et al, 2022) The labelled data is used to train the model, which then makes predictions and learns patterns associated with the labels. Subsequently, the model uses the unlabeled data to improve its understanding of the dataset's underlying structure and relationships.…”

Leveraging Multi-Task Learning to Cope With Poor and Missing Labels of Mammograms

“…The first method is the MixMatch pseudo-labeling technique ( 67 , 68 ), which uses predictions of the pre-trained model as ground truth labels for further training (denoted as “MixMatch”). This method is designed to cope with the lack of labels.…”

Convolutional Networks and Transformers for Mammography Classification: An Experimental Study

“…CNNs have been adopted in many applications related to mammography [ 4 , 12 ], such as mass segmentation [ 13 ], mass detection [ 14 , 15 , 16 ], calcification detection [ 15 , 17 ], mammography classification [ 18 , 19 ], classification of pre-segmented masses [ 20 ]. Most of these works use digitized screen-film mammograms datasets like the Digital Database for Screening Mammography (DDSM) [ 21 ], consisting of 2620 images, or InBreast [ 22 ] which consists of only 410 full-field digital mammograms, or both.…”