A mobile application based on efficient lightweight CNN model for classification of B-ALL cancer from non-cancerous cells: A design and implementation study
Background: B-cell acute lymphoblastic leukemia (B-ALL) is one of the most widespread cancers, and its definitive diagnosis demands invasive and costly diagnostic tests with side effects for patients. Access to definitive diagnostic equipment for B-ALL is limited in many geographical areas. Blood mi...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2023-01-01
|
| Series: | Informatics in Medicine Unlocked |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2352914823000862 |
| _version_ | 1797801012122091520 |
|---|---|
| author | Azamossadat Hosseini Mohammad Amir Eshraghi Tania Taami Hamidreza Sadeghsalehi Zahra Hoseinzadeh Mustafa Ghaderzadeh Mohammad Rafiee |
| author_facet | Azamossadat Hosseini Mohammad Amir Eshraghi Tania Taami Hamidreza Sadeghsalehi Zahra Hoseinzadeh Mustafa Ghaderzadeh Mohammad Rafiee |
| author_sort | Azamossadat Hosseini |
| collection | DOAJ |
| description | Background: B-cell acute lymphoblastic leukemia (B-ALL) is one of the most widespread cancers, and its definitive diagnosis demands invasive and costly diagnostic tests with side effects for patients. Access to definitive diagnostic equipment for B-ALL is limited in many geographical areas. Blood microscopic examination has always been a major B-ALL screening and diagnosis technique. Still, the examination of blood microscopically by laboratory personnel and hematologists is riddled with disadvantages. Meanwhile, AI techniques can achieve remarkable results in blood microscopy image analysis. The present study aimed to design and implement a well-tuned based on deep CNN to detect B-ALL cases from hematogones and then determine the B-ALL subtype. Methods: Based on the well-designed and tuned model, a mobile application was also designed for screening B-ALL from non-B-ALL cases. In the modeling stage, a unique segmentation technique was used for color thresholding in the color LAB space. By applying the K-means clustering algorithm, and adding a mask to the clustered images, a segmented image was obtained to eliminate unnecessary components. After comparing the efficiency of three notable architectures of lightweight CNN (EfficientNetB0, MobileNetV2, and NASNet Mobile), the most efficient model was selected, and the proposed model was accordingly configured and tuned. Results: The proposed model achieved an accuracy of 100%. Finally, a mobile application was designed based on this state-of-the-art model. In the real laboratory setting, the mobile application based on the proposed model classified B-ALL cases from other classes and achieved a sensitivity and specificity of 100% as a robust screening tool. Conclusions: The application that relies on preprocessing and DL algorithms can be used as a powerful screening tool by hematologists and clinical specialists to ignore or minimize unnecessary bone marrow biopsy cases and decrease the B-ALL diagnosis time. |
| first_indexed | 2024-03-13T04:44:59Z |
| format | Article |
| id | doaj.art-f305a5646db043caaca63d1756906a91 |
| institution | Directory Open Access Journal |
| issn | 2352-9148 |
| language | English |
| last_indexed | 2024-03-13T04:44:59Z |
| publishDate | 2023-01-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Informatics in Medicine Unlocked |
| spelling | doaj.art-f305a5646db043caaca63d1756906a912023-06-19T04:28:59ZengElsevierInformatics in Medicine Unlocked2352-91482023-01-0139101244A mobile application based on efficient lightweight CNN model for classification of B-ALL cancer from non-cancerous cells: A design and implementation studyAzamossadat Hosseini0Mohammad Amir Eshraghi1Tania Taami2Hamidreza Sadeghsalehi3Zahra Hoseinzadeh4Mustafa Ghaderzadeh5Mohammad Rafiee6Department of Health Information Technology and Management, School of Allied Medical Sciences, Shahid Beheshti University of Medical Sciences, Tehran, IranSchool of Electrical Engineering, Iran University of Science and Technology, Tehran, IranDepartment of Computer Science, Tallahassee, FL, USADepartment of Neuroscience, Faculty of Advanced Technologies in Medicine, Iran University of Medical Sciences, Tehran, IranPhysical Education Faculty, University of Tabriz, Tabriz, IranDepartment of Artificial Intelligence, Smart University of Medical Science, Tehran, Iran; Corresponding author.Department of Medical Laboratory Sciences, School of Paramedical Sciences, Zanjan University of Medical Science, Zanjan, IranBackground: B-cell acute lymphoblastic leukemia (B-ALL) is one of the most widespread cancers, and its definitive diagnosis demands invasive and costly diagnostic tests with side effects for patients. Access to definitive diagnostic equipment for B-ALL is limited in many geographical areas. Blood microscopic examination has always been a major B-ALL screening and diagnosis technique. Still, the examination of blood microscopically by laboratory personnel and hematologists is riddled with disadvantages. Meanwhile, AI techniques can achieve remarkable results in blood microscopy image analysis. The present study aimed to design and implement a well-tuned based on deep CNN to detect B-ALL cases from hematogones and then determine the B-ALL subtype. Methods: Based on the well-designed and tuned model, a mobile application was also designed for screening B-ALL from non-B-ALL cases. In the modeling stage, a unique segmentation technique was used for color thresholding in the color LAB space. By applying the K-means clustering algorithm, and adding a mask to the clustered images, a segmented image was obtained to eliminate unnecessary components. After comparing the efficiency of three notable architectures of lightweight CNN (EfficientNetB0, MobileNetV2, and NASNet Mobile), the most efficient model was selected, and the proposed model was accordingly configured and tuned. Results: The proposed model achieved an accuracy of 100%. Finally, a mobile application was designed based on this state-of-the-art model. In the real laboratory setting, the mobile application based on the proposed model classified B-ALL cases from other classes and achieved a sensitivity and specificity of 100% as a robust screening tool. Conclusions: The application that relies on preprocessing and DL algorithms can be used as a powerful screening tool by hematologists and clinical specialists to ignore or minimize unnecessary bone marrow biopsy cases and decrease the B-ALL diagnosis time.http://www.sciencedirect.com/science/article/pii/S2352914823000862LightweightMobile applicationAcute lymphoblastic leukemiaefficientnetb0mobilenetv2NASNet mobile |
| spellingShingle | Azamossadat Hosseini Mohammad Amir Eshraghi Tania Taami Hamidreza Sadeghsalehi Zahra Hoseinzadeh Mustafa Ghaderzadeh Mohammad Rafiee A mobile application based on efficient lightweight CNN model for classification of B-ALL cancer from non-cancerous cells: A design and implementation study Informatics in Medicine Unlocked Lightweight Mobile application Acute lymphoblastic leukemia efficientnetb0 mobilenetv2 NASNet mobile |
| title | A mobile application based on efficient lightweight CNN model for classification of B-ALL cancer from non-cancerous cells: A design and implementation study |
| title_full | A mobile application based on efficient lightweight CNN model for classification of B-ALL cancer from non-cancerous cells: A design and implementation study |
| title_fullStr | A mobile application based on efficient lightweight CNN model for classification of B-ALL cancer from non-cancerous cells: A design and implementation study |
| title_full_unstemmed | A mobile application based on efficient lightweight CNN model for classification of B-ALL cancer from non-cancerous cells: A design and implementation study |
| title_short | A mobile application based on efficient lightweight CNN model for classification of B-ALL cancer from non-cancerous cells: A design and implementation study |
| title_sort | mobile application based on efficient lightweight cnn model for classification of b all cancer from non cancerous cells a design and implementation study |
| topic | Lightweight Mobile application Acute lymphoblastic leukemia efficientnetb0 mobilenetv2 NASNet mobile |
| url | http://www.sciencedirect.com/science/article/pii/S2352914823000862 |
| work_keys_str_mv | AT azamossadathosseini amobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy AT mohammadamireshraghi amobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy AT taniataami amobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy AT hamidrezasadeghsalehi amobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy AT zahrahoseinzadeh amobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy AT mustafaghaderzadeh amobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy AT mohammadrafiee amobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy AT azamossadathosseini mobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy AT mohammadamireshraghi mobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy AT taniataami mobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy AT hamidrezasadeghsalehi mobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy AT zahrahoseinzadeh mobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy AT mustafaghaderzadeh mobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy AT mohammadrafiee mobileapplicationbasedonefficientlightweightcnnmodelforclassificationofballcancerfromnoncancerouscellsadesignandimplementationstudy |