Advanced search
Start date

IAssist - Medical Assistant

Grant number: 21/12040-3
Support Opportunities:Research Grants - Innovative Research in Small Business - PIPE
Duration: March 01, 2022 - August 31, 2024
Field of knowledge:Physical Sciences and Mathematics - Computer Science
Convênio/Acordo: SEBRAE-SP
Principal Investigator:André Gustavo Cavalcanti de Melo
Grantee:André Gustavo Cavalcanti de Melo
Host Company:IAssist Desenvolvimento de Programas Ltda
CNAE: Desenvolvimento e licenciamento de programas de computador customizáveis
Atividades de atenção ambulatorial executadas por médicos e odontólogos
Atividades de serviços de complementação diagnóstica e terapêutica
City: São Paulo
Associated researchers:André Fujita ; João Paulo Papa ; Mariangela Correa ; Miriam Galvonas Jasiulionis ; Vladmir Cláudio Cordeiro de Lima
Associated research grant:20/05779-0 - IAssist - Medical Assistant, AP.PIPE
Associated grant(s):21/15044-0 - IAssist - Medical Assistant, AP.PIPE
Associated scholarship(s):22/13262-2 - IAssist - Medical Assistant, BP.TT
22/10923-8 - IAssist - Medical Assistant, BP.TT
22/02474-9 - IAssist: medical assistant, BP.TT
22/01629-9 - IAssist: medical assistant, BP.PIPE


This project aims to continue the activities being developed in Phase 1 under the PIPE-FAPESP Program, process no. 2020/05779-0, effective from 04/01/2021 to 12/31/2021."IAssist - Medical Assistant" is an application based on AI and Machine Learning for use by the medical community that provides primary and secondary care to patients, especially for those professionals located in regions far from reference centers, with a lack of specialist doctors Oncology. The solution can be classified as a Clinical Decision Support System (CDSS), aimed at assisting the physician in the cancer diagnosis process. The delay in diagnosis and the shortage of specialists are among the main factors that contribute to the increase in cancer mortality rates in Brazil. Greater assertiveness in identifying the patient's condition minimizes the possibility of diagnostic errors and gives the physician more security in referring the patient to a secondary or tertiary health service (specialized oncology center), thus reducing the time to start the treatment, providing a better quality of life, improving patient survival and allowing the reduction of costs related to unnecessary consultations and exams, thus reducing the pressure on public and private health services. The tool proposed here is presented as a viable and effective solution to aid in medical diagnosis. Based on algorithms that are learning to relate signs, symptoms and clinical data of patients from anonymized data from patient records from cancer reference centers, with which we have established Terms of Technical Cooperation, such as the Hospital de Amor (Hospital de Cancer de Barretos) and Hospital de Cancer Infanto-juvenil de Brasília José de Alencar (controlled environments), as well as non-oncological hospitals (uncontrolled environments), the application processes clinical and laboratory data of patients, in order to offer to doctors who provide primary and secondary care to a patient the suggestion of a probable cancer diagnosis and the percentage of assertiveness of this suggestion. In this initial Phase 1, the analyzes of 2,000 medical records of non-oncological infant-juvenile patients with central nervous system or abdominal cavity tumors by the developed algorithms revealed a high degree of accuracy, sensitivity and specificity, confirming the feasibility of the proposal. In Phase 2 of this project, we will expand the analysis of juvenile tumors with the highest incidence. The solution can also be used to periodically run-in batch mode a large volume of data from patient records of contracting health institutions, seeking to identify signs and symptoms that may indicate possible cancer cases, thus favoring early diagnosis. Digital tools that aid clinical decision making (CDSS) have been around for some time, but are mainly based on image analysis or molecular markers. The factors that characterize the innovation of the solution proposed in this project are: (1) data entries can be performed with unstructured data in natural language, exactly as shown in the patient's medical record; (2) can be fed directly into the app or by integration with digital health platforms; and (3) the entire process is carried out in Portuguese. Despite extensive research, we have not identified any direct competitors, in Brazil or abroad, that present these characteristics. Among the target customers of this solution are the Basic Health Units (UBS) of the 5,568 Brazilian municipalities, other public and private health services, health care plan operators and digital health platforms. (AU)

Articles published in Agência FAPESP Newsletter about the research grant:
Articles published in other media outlets (0 total):
More itemsLess items

Please report errors in scientific publications list by writing to: