1. General

Category

Category

Health

2. Project Details

Company or Institution

AI Care Private Limited

Project

Tashkhees

General description of the AI solution

AI-based solutions have machine learning / deep learning techniques underneath that model the human brain for automating day to day decision making that would otherwise take hundreds to thousands of human hours to achieve. Incorporating human level learning capability into the systems require development of extensive application specific tagged datasets, which fulfil the ethical and scientific criteria. That are then fed into the mathematical models that crunch the data and see patterns that help in taking real-world decisions to solve the problems at hand.
Since AI is data driven, the efficacy of AI-based solutions is dependent on the quality of data sets fed into them. Normally the data is divided into three subsets: training; validation; and test. The training data, as the name suggests, is the key to training the system. Training times are quite long depending on the data. Training can be further improved by setting the hyper-parameters of the learning system and is done during the validation phase. Finally, system’s performance is evaluated on the unseen test data, which is different from the training set.
An important aspect of the AI-based solutions is interpretability, which explains how the factors which contribute towards particular decisions taken by the system. That makes the AI algorithms more transparent which otherwise are more like black boxes. That is crucial for the practitioners in the real world to validate decision making process of the AI-based systems and enhances trust of the people in it.

Website

https://aicaresol.com/

Organisation

Medical Imaging and Diagnostics Lab COMSATS University Isamabad

3. Aspects

Excellence and Scientific Quality: Please detail the improvements made by the nominee or the nominees’ team or yourself if your applying for the award, and why they have been a success.

Our AI solution is addressing third SDG of Health and wellbeing and being a custodian of technology solutions, our scope of interest is to diagnose cancerous diseases in less time with a minimal cost involved.
Despite all the advances’ made cancer is still one of the leading causes of death in Pakistan and other region of the world. Specifically, every year around 40,000 women succumb to breast cancer. Anecdotal evidence as narrated by radiologists suggests that recently, more younger women are found to have the disease which is alarming. Brain Tumor, CNS tumors in general, is the second most common cause of death among younger population less than 15 years. One of the major causes of high mortality is the late detection of cancer, mostly in stage (III/IV) even in advanced health care systems.
At that stage they are relatively much harder to treat. Thus, early detection is the key to saving lives and it forms the keystone of this project. In this project, we aim to commercialize Genomics, Histopathology, MRI, CT and X-ray based artificial intelligence tool for cancer detection which will impact healthcare sector with the market value of almost 2-3 million USD (https://www.webfx.com/internet-marketing/ai-pricing.html), and also provide gene sequencing services. With near to 200,000 cancer diagnoses each year in Pakistan alone, potential size for the local market is big. And this solution can easily be deployed in public cancer treating hospitals with potential to deploy them at private hospitals. As per our preliminary working, more than ten plus public and private sector hospitals can come under the umbrella of this solution.
Artificial Intelligence is creating boom everywhere and sprouting day by day. The growth of AI is never ending. It has marked its impressions in every field of life and has left no stone unturned. The magnificent involvement and success of AI cannot be overlooked in the field of healthcare industry. Cancer is the biggest medical problem concerning life and is considered as the deadliest disease. Brain tumors is considered as the most complex tumors, difficult to detect because of the heterogenous nature and varying shape/size. We have made an AI based diagnostic tool to help assist the radiologist for its early diagnosis and prognosis.
The quality of our AI based solution can be considered from the fact that the model behind this solution stood third in an international BRATS challenge which held every year since 2012. the model is complete and takes 3D MRI images as input and gives the segmentation as well as the survival prediction of the brain tumor patient in days.
Well, our products are AI based solutions to medical problems so we mainly use technical metrics like accuracy to check the confidence level of our products. All of our products have accuracy higher than 90%. Since our establishment, we have attended numerous conferences, published around 15 research papers and participated in many national and international challenges and workshops. As already stated above, we won the third position in BRaTS challenge 2020 in survival prediction task. currently we have held the IP rights of all of our products and we are ready to take them to the next level through some joint collaborative project.

Scaling of impact to SDGs: Please detail how many citizens/communities and/or researchers/businesses this has had or can have a positive impact on, including particular groups where applicable and to what extent.

Our AI solution is addressing third SDG of Health and wellbeing and being a custodian of technology solutions, our scope of interest is to diagnose cancerous diseases in less time with a minimal cost involved.
Despite all the advances’ made cancer is still one of the leading causes of death in Pakistan and other region of the world. Specifically, every year around 40,000 women succumb to breast cancer. Anecdotal evidence as narrated by radiologists suggests that recently, more younger women are found to have the disease which is alarming. Brain Tumor, CNS tumors in general, is the second most common cause of death among younger population less than 15 years. One of the major causes of high mortality is the late detection of cancer, mostly in stage (III/IV) even in advanced health care systems.
At that stage they are relatively much harder to treat. Thus, early detection is the key to saving lives and it forms the keystone of this project. In this project, we aim to commercialize Genomics, Histopathology, MRI, CT and X-ray based artificial intelligence tool for cancer detection which will impact healthcare sector with the market value of almost 2-3 million USD (https://www.webfx.com/internet-marketing/ai-pricing.html), and also provide gene sequencing services. With near to 200,000 cancer diagnoses each year in Pakistan alone, potential size for the local market is big. And this solution can easily be deployed in public cancer treating hospitals with potential to deploy them at private hospitals. As per our preliminary working, more than ten plus public and private sector hospitals can come under the umbrella of this solution.

Scaling of AI solution: Please detail what proof of concept or implementations can you show now in terms of its efficacy and how the solution can be scaled to provide a global impact ad how realistic that scaling is.

AI is substantially faster at image analysis and enables the automation of manual, time-consuming tasks. Speeding up case review increases your Radiology output allowing for more new patients to be taken in. Additionally with the time saved, radiologists can focus for longer on complex and rare cases. Acquiring data sets are the first major task and sustainability of a solution is completely depending upon the accuracy. Research phase of a product development required immense amount of time and cost in order to get faster results. One of the major problems in getting more accurate results is to get the annotated images from radiologists, to achieve more accuracy in real time testing.
Improved diagnostic accuracy is achieved with AI-assistance as bias and subjectivity are eliminated. AI systems analyze cases with 100% consistency. Reducing diagnostic error and misdiagnosis, while improving treatment accuracy with more detailed results, will result in direct cost savings due to greater precision. As per SDG of good health and wellbeing AI diagnostics not only save cost but it also saves life through early detection. Our solution is a tool for radiologists which can handle bulk tasks for radiologists which will help them to get the required results in seconds and this can generate reports also which is an output of radiology task. The improved diagnostic accuracy and consistency of analysis with AI assistance brings benefits not only to the hospital and radiologists but perhaps more importantly, to the patients as well. Collaboration is first major step to adopt technologies and most of the research centers and universities are doing research in AI diagnostics. Our research goals and studies will not only benefit the global ecosystem to adopt solutions but it can generate revenues through refined models of business. Our user community is big and adoption rate is slow due to more relaying on manual diagnostics, recent boom of industry 4.0 have enhanced the usage of automation techniques in every field of research and development.

Ethical aspect: Please detail the way the solution addresses any of the main ethical aspects, including trustworthiness, bias, gender issues, etc.

As is well known that AI based solutions themselves are not compliant with ethical standards and have certain biases built-in into the datasets. To resolve such issues a pro-active approach on the part of the designers of the AI systems is required whereby they ensure compliance of the system with the regulatory regime and also to actively collect and curate data so as to reduce to the minimum the ethical issues that may arise and also to reduce biases within the data.
In our field of work, ethical standards require that privacy of the patients is maintained. Anonymization of data is a standard practice whereby the private information which is normally not of interest to the AI people is not considered and is ignored, barring cases where such information may be useful for diagnostic purposes. For instance, the age of a patient is normally ignored when the aim is to identify the disease in the radiology images but is of utmost importance when prognosis is to be done for the patient as older patients may have lesser chances of survival after diagnosis. But to ensure compliance with ethical standards patients may be given a choice for their case to be made part of the research study. Their cases won’t be considered if they do not consent to it. It is pretty much standard practice in the domain.
AI systems are hungry for data and they are only as good as the quality of data fed into them. Thus, data curation is the most important aspect of developing an AI based system. But data in the real world suffers from biases. If a greater portion of blacks are in jail for various crimes, the AI system may learn to classify images or videos of blacks into criminal category. Similarly, if the data of a certain disease carry more cases of people belonging to a particular racial, having a particular gender, or belonging to a particular socio-economic background, the propensity of the AI system to classify them as carrying the disease would be high. To offset the effect of such biases, special care should be taken by the curators to ensure that people belonging to one particular background are not overly represented in the data, and all have equal representation.