An radiologist scans a medical image on an average 1.5 seconds with an observed error rate 3-5%. Such a lacunae in medical services results in mis-diagnosis of diseases. For example, it has been reported that COPD is missed during diagnosis almost five times the correctly diagnosed COPD. We are developing an AI tool (with API) that classifies respiratory diseases such as cardiomegaly, pneumonia, etc . As a first step our AI tool identifies Cardiomegaly with AUC of 0.93 and pneumonia with AUC of 0.97. This metric is an improvement over current state of art.

Clinical Technicians, who are the first point of contact in a remote rural setup, are tasked with identification of these respiratory disease. Since its heavily dependant on technician’s experience, an element of human error is involved, and consequently there are chances of disease being overlooked or misdiagnosed. Also, during night shifts expert radiologist might not be present. In both the above scenarios, the problem can be ...
An radiologist scans a medical image on an average 1.5 seconds with an observed error rate 3-5%. Such a lacunae in medical services results in mis-diagnosis of diseases. For example, it has been reported that COPD is missed during diagnosis almost five times the correctly diagnosed COPD. We are developing an AI tool (with API) that classifies respiratory diseases such as cardiomegaly, pneumonia, etc . As a first step our AI tool identifies Cardiomegaly with AUC of 0.93 and pneumonia with AUC of 0.97. This metric is an improvement over current state of art.

Clinical Technicians, who are the first point of contact in a remote rural setup, are tasked with identification of these respiratory disease. Since its heavily dependant on technician’s experience, an element of human error is involved, and consequently there are chances of disease being overlooked or misdiagnosed. Also, during night shifts expert radiologist might not be present. In both the above scenarios, the problem can be solved using an AI software that predicts diseases from X ray scans.

Another problem that we intend to solve is the problem of
high turnaround time in time sensitive, high volume trauma centers. Every 21 seconds, one person in the U.S. sustains a brain injury and brain injury contributes to about 30% of all injury deaths. In trauma centers such as brain trauma center, due to human component involved there exist a time delay between when the doctor requests a CT Scan to the time he receives the CT scan report along with radiologist report. This delay in diagnosis and treatment impinges upon the ‘golden hour’ and reduces the survival rate of the patient. Moreover due to human component and intense environment of the trauma center, there is a significant probability of misdiagnosis. An AI tool that predicts the severity of brain injury would reduce the delay in triage in patient’s prognosis.

In all of the above scenarios, we foresee that our product Doctor’s AI assistant toolbox would augments clinicians capabilities by being deployed independently as AI software and/or the AI API could be integrated with medical devices such as X-ray machine, CT Scan, etc .
More information

Employees

Priyanka Nath
Admin
Priyanka Nath CEO Focused, product-oriented data Scientist with knack for solving problems