There are many Indian startups claiming to develop artificial intelligence software to help the healthcare industry. However, their websites are filled with marketing terms and hype that could lead one to believe a company leverages AI when it does not. In sectors with a booming landscape of startups, we caution readers to take a closer look at each company before considering any of their solutions.
We have found that AI-powered solutions for healthcare in India can be found across almost all AI applications. Machine vision is leveraged for medical imaging, while predictive analytics is used to find insights about patient health in more specific areas such as cardiology and pulmonology. Both are used in diagnostics. While natural language processing (NLP) can be found all over the healthcare industry in other parts of the world, we only found one Indian company using it for their solutions.
Unlike the more white collar automation NLP solutions typically offer, we found more companies leveraging AI for specific medical procedures. It would appear that innovation in the hands-on diagnostic and preventative space of the healthcare industry is what is mainly driving success for this surge of startups.
Sigtuple is offering software called Shonit, which can purportedly upload blood smear images straight from the microscope to the cloud for cell counting and visual anomaly detection. The cloud AI platform allows every blood smear to help the machine learning model learn and for pathologists to access this data remotely. Sigtuple calls this telepathology and advertises it as an important use for Shonit.
Tricog’s Insta ECG software also allows for the telecommunication of medical data and is able to produce detailed ECG reports showing anomalies in heartbeat. Their software is embedded in their ECG report printing machine from General Electric.
Mapmygenome’s Genomepatri system allows them to take DNA samples from customers to perform AI-powered analytics on their genes and find causes for medical diseases or disorders. The company uses these analyses to diagnose the customer and find them the best course of action for seeking medical attention and keeping healthy habits.
Zerone Consulting is offering NLP software to help with clinical documentation accuracy and document search. They claim their solutions can improve their client’s ability to quickly analyze large amounts of unstructured text data.
The companies we found in India claiming to offer AI-powered solutions to healthcare providers advertise their software to handle at least one of the following problems:
- Medical Imaging and Telepathology
- ECG Testing and Emergency Management
- Genetic Testing
- EHR Data Search and Summarization
We’ll begin this report with a discussion of Sigtuple, a startup offering AI software for medical imaging and telepathology.
Medical Imaging and Telepathology
Sigtuple offers a blood smear analyzer software called Shonit which it claims can help pathologists diagnose illnesses and conditions like anemia using machine vision.
Shonit is a cloud-based machine learning solution that enables telepathology for its clients. Microscoping blood smear images of patients are uploaded to the cloud for analysis and are accessible by pathologists who may not be on site. Sigtuple achieves this with their smart microscope. The smart microscope is a traditional microscope fitted with a Shonit-enabled smartphone.
The smartphone’s camera takes pictures through the microscope lenses and sends those pictures to the cloud AI platform which detects cell counts and abnormalities in the blood. Sigtuple claims these illnesses could range from anemia to blood cancer.
The machine learning model behind the Shonite software was likely trained on hundreds of thousands of microscopic images of blood smears. Each image would need to show blood with different cell counts, illnesses, or conditions.
These images would need to be labeled according to their cell counts and which illnesses they might indicate. The labeled blood smears would then be run through Shonit’s machine learning algorithm. This would train the algorithm to discern the patterns of 1’s and 0’s that appear to humans as a microscopic image of a patient’s blood as displayed in a blood smear image.
The user could then observe a new blood smear that is not labeled under the Shonit smart microscope. The algorithm behind the software would then be able to indicate how many cells are present in the blood smear and any illnesses that the smear might indicate. The system would then alert the user of these metrics and upload them into the cloud.
Below is a short 3-minute video explaining how Sigtuple’s Shonit software scans blood smears, sends them to the cloud for analysis, and then to pathologists for diagnosis:
Sigtuple‘s software is still in a partner-exclusive beta launch, so they do not have any case studies showing a healthcare company’s success with the software.
Sigtuple does not list any major companies as clients, but they have raised $24.8 million and are backed by Trifecta Capital and Accel.
Tathagato Rai Dastidar is Co-founder and Chief Scientific Officer at Sigtuple. He holds a PhD in computer science and engineering from the Indian Institute of Technology. Previously, Dastidar served as Senior Director of R&D at Tribune Digital Ventures.
ECG Testing and Emergency Management
Tricog Health offers software called Insta ECG, which it claims can help healthcare providers create accurate cardiology reports and prevent heart attack mortalities using what appears to be predictive analytics.
Insta ECG requires specific hardware offered by Tricog in order to work. The main device used to create the ECG reports is the General Electric MAC 600, which is fitted with a Tricog communicator. The device comes with the necessary cords and body connectors for installation. These include cuff shaped sensors for the patient’s ankles and wrists, as well as six sensors that are equipped to the chest.
Once the machine learning model creates the ECG report, it is both printed and sent to Tricog Health’s database, where it can be accessed by their employees should a client call them for support. Tricog offers analysis of these reports by their specialists as well as a cardiology team who answers questions about information or anomalies found in the reports.
The company claims their specialists are able to help in times of critical diagnoses, working with clients to help prevent and treat heart attacks and raise the patient’s survival rate.
We can infer the machine learning model behind Insta ECG was trained on tens of thousands of ECG tests and records holding cardiological data points. These data points would involve heart rhythm, T waves and other vibrations given off by the heartbeat, or anomalies within those. An example of this would be an inversion of T waves as they come through the ECG machine.
A data scientist would then run the data through the software’s machine learning algorithm. This would train the algorithm to discern which data points correlate to abnormalities in heart activity. The anomalies would be accurately depicted in the ECG record printed from the Tricog hardware and sent to Tricog’s specialists digitally.
The software would then be able to produce an ECG report showing a patient’s heartbeat labeled to be read by a cardiologist. This report would reveal any anomalies, as well as contain information such as heartbeats per minute (bpm). However, this could require the physician to upload information about the patient’s recent cardiological history or new medications beforehand.
It’s likely that data scientists can integrate the software into healthcare databases where ECG and other patient cardiology data is stored.
Below is a short video explaining the four-step process of how Insta ECG is used. Tricog lists them as the following:
- Digital Data Transfer
- ECG Diagnosed by Experts
- Report sent in six minutes
Tricog Health does not make available any case studies showing a healthcare company’s success with the software.
Tricog Health does list Manasa Trinity Heart Hospital, Indian Institute of Science as some of their past clients.
Zainul Charbiwala is Co-founder and CTO at Tricog Health. He holds a PhD in Electrical Engineering and Circuits and Embedded Systems from UCLA. Previously, Charbiwala served as Research Manager at IBM India.
Mapmygenome offers a proprietary software called Genomepatri, which it claims can help customers make an action plan for their personal health based on their genetics using predictive analytics. Because the software is mainly used for creating patient-specific health suggestions, we could more accurately categorize this as prescriptive analytics.
The Genomepatri machine learning model would need to be trained on thousands of DNA samples from patients or volunteers containing genetic data. Individual genes can be analyzed to find how they affect body growth and development of illnesses, so a single sample of genetic data can be used to find numerous potential issues involving all parts of the body.
A data scientist would have then exposed this data to the software’s machine learning algorithm. This would have trained the algorithm to discern which data points correlate faulty genes to specific diseases or disorders.
The software would then be able to make predictions about a customer’s condition or what that condition might be. This helps find accurate diagnoses and related statistics, such as the chance symptoms like hair loss will stop or get better.
Below is a screencap from Mapmygenome.in showing the process by which the company collects, analyzes, and reports on customers’ DNA:
Mapmygenome claims to have helped an individual suffering from total loss of head and body hair. The 30-year-old patient was diagnosed with Alopecia Universalis, most likely triggered from stress or an abnormality with his immune system. He wanted to know if his hair loss was genetic, if there was any chance of growing his hair back, or if he could pass this condition on to his children.
According to the case study, Genomepatri found a nonworking copy of a gene that is commonly associated with Clouston Syndrome. This informed the analysis used to explain his 50% chance to pass on the condition to his children. Eye-related complications are common in those with Clouston Syndrome. Mapmygenome referred the patient to an ophthalmologist for further screening.
Mapmygenome lists Apollo Hospitals and Manipal Hospitals as some of their past clients.
Anu Acharya is CEO at Mapmygenome. She holds an MS in Physics from the University of Illinois: Chicago. Previously, Acharya served as CEO at Ocimum Biosolutions.
EHR Data Search and Summarization
Zerone Consulting offers a namesake software which it claims can help healthcare companies simplify medical documentation and summarize data from large amounts of documents using natural language processing.
The software appears to be for use with electronic health records (EHR) and other clinical documentation. In addition, the company claims the software can comb through large amounts of text to find important insights, which also analyzes the semantic meaning of phrases to find further information. For example, the software could comb over two sentences that help to explain the same concept and find a better explanation from the data extracted from both.
The company states their machine learning model for NLP would need to be trained on their client’s text data they want to search and analyze more easily. For the healthcare industry, this could consist of tens of thousands of EHRs, patient profiles, prescription, bills, and invoices. These documents would be labeled as the type of document they are, and important keywords, phrases, and health statistics within the text would be tagged for search results. A data scientist would then expose Zerone Consulting’s machine learning model to this data. This would train the algorithm to discern the chains of text that to humans might appear as an EHR, patient profile, prescription, bill, or invoice.
The software would then be able to determine which keywords correlate to which documents and organize data points extracted from text to optimize searching for users. However, this could require the user to upload information about recent updates to certain medical documents or future changes to healthcare procedures into the software prior to running the software.
We could not find a demonstration video showing how Zerone Consulting’s NLP software works. In addition, they do not make available any images or graphics organizers explaining how their software works.
Zerone Consulting claims to have helped a leading provider of healthcare procurement services automate their invoice verification and identification process. The client company integrated Zerone Consulting’s software into its existing database of incoming invoices, as well as the current OCR solution from ABBYY. That solution was likely ABBYY’s FlexiCapture software; however, this was unclear. Zerone Consulting claims their machine learning model helped their client optimize their ABBYY software through training and customization for invoice processing. According to the case study, the client company was able to optimize their invoice verification cycle and saw 50% decrease in the cost of each invoice.
Zerone Consulting does not list any of their past clients by name on their website.
Sreejith Madhavan is Chief Operating Officer at Zerone Consulting. He holds a Master of Computer Applications (MCA) degree in Computer Science from Bharathiar University. Previously, Madhavan served as a Web Developer at Software Associates.
Header Image Credit: TechStory