Artificial Intelligence Technologies for Multimodal Big Data Analysis for Breast Cancer Diagnosis and Prognosis

Project manager: Abdikenov Beibit Bolatgazyevich

Funding source: PAF KS MSHE OF RK

Goal: Improving breast cancer diagnostics by developing artificial intelligence technologies with subsequent implementation in clinical practice

Implementation years: 2024–2026

Amount of funding: 1,017,720,283 tenge

Partners:

JSC “Kazakh Research Institute of Oncology and Radiology”, RSE on the Right of Economic Management “Institute of Genetics and Physiology” of the Science Committee of the MNVO RK

RSE on the Right of Economic Management “National Scientific Center for Healthcare Development named after Salidat Kairbekova” of the MH RK

KGP on the Right of Economic Management “Almaty Oncology Center” of the Public Health Department of Almaty

Project objectives:

Research will be conducted on regulatory documents for collecting and storing large multimodal (clinical data, biomarkers, text data, images, and genetic data) data;
A database of large multimodal (clinical data, biomarkers, text data, images, and genetic data) data will be prepared using the methodology for collecting, anonymizing, and labeling data;
Algorithms will be developed for the secure storage of structured and unstructured large multimodal (clinical data, biomarkers, text data, images, and genomic data) data;
Deep neural network algorithms will be developed for breast cancer diagnostics based on mammography data.

Матрица реализации проекта

Expected results

Tasks (WP- work packages)	Expected results
Tasks (WP- work packages)
WP1: Development and training of models.	Training of deep neural network models on open-source data. Deep neural network models will be trained. Preparation and submission of articles on deep neural networks. At least two articles and (or) reviews will be published in peer-reviewed scientific journals in the scientific area of the program, included in the 1st (first), 2nd (second) quartile by impact factor in the Web of Science database and (or) having a CiteScore percentile in the Scopus database of at least 50 (fifty) or in journals recommended by the KOKNVO. Development of pre-trained and partially trained Vision-Language models. Vision-Language models will be developed and trained
WP2: Data collection	Survey and data collection. To be completed by the end of 2025: Survey of 446 verified patients for prospective studies; Provision of mammographic images for 446 verified patients; Provision of ultrasound data for 446 verified patients; Provision of magnetic resonance imaging data for 446 verified patients; Digitization of histological slides for 446 verified patients; Provision of digitized data on histological slides for 446 verified patients; Digitization of histological slides for 2,000 verified patients for retrospective studies; Provision of digitized data on histological slides for 2,000 verified patients for retrospective studies; Provision of tabular, text and biomarker data on 446 verified patients; Provision of histological slides for genomic studies on 446 verified patients. To be completed by the end of 2026: Questionnaire survey of 554 verified patients for prospective studies; Provision of mammographic images on 554 verified patients; Provision of ultrasound data on 554 verified patients; Provision of magnetic resonance imaging data on 554 verified patients; Digitization of histological slides on 554 verified patients; Provision of digitized data on histological slides on 554 verified patients; Provision of tabular, text and biomarker data on 554 verified patients; Provision of histological slides for genomic studies on 554 verified patients. Data processing and labeling. Data will be anonymized, processed and labeled for retrospective studies (mammography – 5000, ultrasound – 2000, magnetic resonance imaging – 2000, histology – 2000) by the end of 2025. The following will be completed by the end of 2026: Data anonymization, processing and labeling for retrospective studies (computer tomography – 2000, tabular data – 5000, text data – 2000, biomarkers – 2000); Anonymization, processing and labeling of data for prospective studies (mammography – 1000, ultrasound – 1000, magnetic resonance imaging – 1000, histology – 1000, tabular data – 1000, text data – 1000, biomarkers – 1000). Study and identification of genetic and molecular markers of breast cancer. To be carried out by the end of 2025: Transportation of biomaterials (blood and tissue) of 235 patients; Sequencing of biomaterials of 235 patients Purchase of reagents for preparation and enrichment of libraries, DNA quality control, sequencing for 235 patients Purchase of consumables necessary for the study for 235 patients Bioinformatics analysis of sequencing data of 235 patients Preparation of a scientific article; Preparation of regulatory documents. Will be carried out by the end of 2026: Transportation of biomaterials (blood and tissue) of 265 patients; Sequencing of biomaterials of 265 patients. Purchase of reagents for library preparation and enrichment, DNA quality control, sequencing for 265 patients Purchase of consumables necessary for conducting research for 265 patients Bioinformatics analysis of sequencing data of 265 patients
WP3: Regulatory documents	Conducting research on regulatory documents for collecting and storing large multimodal data. The following will be conducted: Studying and analyzing regulatory documents; Conducting a literature review; Preparing a scientific article; Preparing regulatory documents. Conducting research on regulatory documents for testing the results of artificial intelligence technologies. The following will be conducted: Studying and analyzing regulatory documents; Preparing regulatory documents. Conducting research on regulatory documents for applying the results of artificial intelligence technologies in decision-making by medical workers. The following will be conducted: Studying and analyzing regulatory documents; Conducting a literature review; Preparing a scientific article; Preparing regulatory documents.

Breast Cancer Diagnostics with AI Integration

Results:

research was conducted on regulatory documents for collecting and storing large multimodal data;
a database of large multimodal data was prepared using the methodology of collecting, anonymizing and labeling data;
algorithms were developed for the secure storage of structured and unstructured large multimodal data;
deep neural network algorithms were developed for breast cancer diagnostics based on mammography data

Project Team

research was conducted on regulatory documents for collecting and storing large multimodal data;
a database of large multimodal data was prepared using the methodology of collecting, anonymizing and labeling data;
algorithms were developed for the secure storage of structured and unstructured large multimodal data;
deep neural network algorithms were developed for breast cancer diagnostics based on mammography data