An automatic breast computer‐aided diagnosis scheme based on a weighted fusion of relevant features and a deep CNN classifier

Abstract

Mammography continues to play a central part in the early breast masses diagnosis and has raised several challenges in breast cancer detection. Nevertheless, it remains still difficult to detect abnormalities in a dense breast and some benign lesions may have similar appearance with masses. This study proposes a novel Computer-Aided Diagnosis (CADx) allowing benign and malignant mass classification. Accordingly, relevant textural-shape descriptors are proposed, which is the aggregation of a new Local Binary based feature, namely the Monogenic Gray Level and Local Difference (MGLLD), where both texture characteristics and breast tissue density information, and the Zernike moments are incorporated. A heuristic algorithm is then devised for optimizing the proposed features with respect to inter-class and intra-class distribution, by a convenient ponderation. Then, a set of classifiers are applied for opting to the best decision making solution. The Deep CNN yields the best accuracy, with an Area Under Curve (AUC) of 0.98 on Curated Breast Imaging Subset of DDSM (CBIS-DDSM). In addition, the authors are based on a subjective approval of the extracted Region Of Interest (ROI) by experts in radiology. The proposed scheme proves its effectiveness especially on some challenging breast cancer cases, corresponding to higher breast tissue density. In medical image processing, these results open new perspectives with respect to the use of the deep learning solutions, where the small sample-number is known to be a limitation (because pathologies cannot be intentionally provoked).

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

“…This innovation incorporates several independent paths inside a block using grouped convolutions. (Gnanasekaran et al, 2020;Joseph et al, 2024;Tsai et al, 2022), (Guan and Loew, 2019;Wang and Yang, 2018;Arputham et al, 2021;Chugh et al, 2022;Al-Antari et al, 2018a;Kurek et al, 2018;Thirunavukkarasu et al, 2022;Albalawi et al, 2022;Gargouri et al, 2022;Karthiga et al, 2022;Jafarzadeh Ghoushchi et al, 2021;AlGhamdi and Abdel-Mottaleb, 2021;Oyelade and Ezugwu, 2022;Kumar and Gudur, 2023;Diniz et al, 2018;El Houby and Yassin, 2021;Guan et al, 2020;Tavakoli et al, 2019;Altan, 2020;Kumar Singh et al, 2022;Duggento et al, 2019;Deb et al, 2023;Gao et al, 2020;Houssein et al, 2022;Patil and Biradar, 2021;Sakthivel et al, 2022a;Raaj, 2023;Sridevi and Samath, 2023;Alfifi et al, 2020;Arora et al, 2020;Ranjbarzadeh et al, 2023a;Wang et al, 2021c;Wimmer et al, 2021;Wu et al, 2019;Zhang et al, 2021bZhang et al, , 2018 It is a custom architecture for the extraction and classification of features. Convolutional layers are typically used to capture spatial features, pooling layers to downsample, and fully conne...…”

“…From these tables we can infer that a customized CNN is the most common approach used for the classification of breast density. (Gargouri et al, 2022;Malathi and Latha, 2023;Dafni Rose et al, 2022;Saber et al, 2023;Abdelhafiz et al, 2020) (Patil and Biradar, 2021;Patil et al, 2022;Ragab et al, 2021) Beginning with seed points, it gradually expands regions by incorporating neighboring pixels that share specific similarity criteria, like intensity or texture. The method continues until the termination conditions are satisfied, which ensures that the region reaches the desired size or meets a boundary.…”

“…Inception-v3 improved the model by using factorised convolutions and dimension reductions. (Babu and Jerome, 2022;Mahmood et al, 2022;Shen et al, 2020;Wang et al, 2019;Zhang et al, 2021a), (Albalawi et al, 2022;Cai et al, 2021;Chandraraju and Jeyaprakash, 2022;Gargouri et al, 2022;George Melekoodappattu et al, 2022;Jones et al, 2023;Kurek et al, 2018;Mahmoud et al, 2021;Mokni and Haoues, 2022;Patil and Biradar, 2021;Patil et al, 2022;Razali et al, 2023a;Sajid et al, 2023;Shivhare and Saxena, 2022;Arputham et al, 2021;Chouhan et al, 2021;Chugh et al, 2022;Ramesh et al, 2022;Sivakrithika and Dinakaran, 2018;Ranjbarzadeh et al, 2023a;Song et al, 2020;Yu et al, 2023c;Elmoufidi, 2022;Hosni Mahmoud et al, 2022) (Babu andJerome, 2022), (Razali et al, 2023a;Arputham et al, 2021;Chouhan et al, 2021) (Mahmood et al, 2022;Shen et al, 2020;Wang et al, 2019;Zhang et al, 2021a), (Albalawi et al, 2022;Razali e...…”