Medical Imaging and Diagnostics Lab


Principal Investigator:

Dr. Ahmad Raza Shahid

Department of Computer Science, COMSATS University, Islamabad

Park Road, Tarlai Kalan, Islamabad 45550, Pakistan

Phone: +92-51-90495337, Mobile: +92-334-8633910, Email:


CO-Principal Investigators:

Brain Tumor Detection: Dr. Tahir Mustafa Madani

Department of Computer Science, COMSATS University, Islamabad

Park Road, Tarlai Kalan, Islamabad 45550, Pakistan

Phone: +92-345-5190120, Email:


Breast Cancer Detection: Dr. Amir Hanif Dar

Department of Computer Science, COMSATS University, Islamabad

Park Road, Tarlai Kalan, Islamabad 45550, Pakistan

Phone: +92-305-5147377, Email:


Tuberculosis Identification: Dr. Tehseen Zia

Department of Computer Science, COMSATS University, Islamabad

Park Road, Tarlai Kalan, Islamabad 45550, Pakistan

Phone: +92-321-6056095, Email:


Goal: Use of AI based techniques and approaches on different image modalities, such as MRI, CT-Scan, X-ray, and Ultrasound to identify different diseases such as cancer and tuberculosis.

Scientific Objectives:

CAD systems cover a wide array of diseases and image modalities. Mainly we will focus on the following diseases and image modalities:

  • Use of mammograms for breast cancer detection
  • Use of chest X-ray for tuberculosis (TB) identification
  • Use of MRI for brain tumor detection

The reason for selecting the diseases and image modalities is that they occur in high numbers in our society and the chosen image modalities are helpful in detecting the earlier presence of the disease, which may be invisible to the naked eye.

While brain and other Central Nervous System (CNS) tumors are the 10th leading cause of death, around 40,000 women in Pakistan succumb to breast cancer each year, and 420,000 new cases of TB are diagnosed in Pakistan each year. The cases where the disease is discovered at a later stage put extra burden on the national exchequer and affects the patients, their friends and relatives both emotionally and financially.

One may opt for importing complete solutions from abroad, but they have a higher price tag and does not help in indigenous development of the technology and know-how. Thus, there is a dire need to develop such technology locally, at a cheaper price that caters to specific symptoms and diseases in the local context.

The lab aspires four scientific goals:

a) Solve local problems using AI-based techniques and take solutions to market through technology commercialization and licensing.

Value to the country: This goal targets multifaceted achievements. For one, local problems are eradicated through technology-based solutions. Secondly, technology commercialization will spur new employment opportunities and will also stimulate new business and revenue streams.

b) Provide high-value services to academia and industrial partners.

Value to the country: Academia and industry would be able to profit from the expertise developed in the wake of the execution of the proposed projects. The computationally heavy physical resources that will be made available will make use of latest techniques possible for our researchers, who are normally handicapped by the absence of such resources.

c) Develop an advanced workforce in medical image processing through trainings and applied work.

Value to the country: The lab will not just be a research center but will also provide training to students working towards their MS and PhD theses. It will also start new MS and  PhD programs in Artificial Intelligence, which would focus on both the breadth and the depth of the area.

d) Become a catalyst for sustainable economic growth.

Value to the country: Highly ranked foreign universities contribute in the social and economic fabric of the society. Unfortunately, Pakistani universities mostly invest in teaching and producing research papers. The planned lab, in contrast, shall become an archetype of an academic institute plus center of medical imaging and diagnostics research, contributing in economic growth, by achieving the above three goals. It shall inspire other institutes and labs to follow the footsteps and thus create a visible impact at the national scale.


Description of Lab:

With the advent of medical imaging technology, digital graphical representation of different parts of the body are readily available in different modalities, such as MRI, CT-Scan, and X-ray. That, along with the development in computer hardware technology, has made it possible for computer scientists to apply machine learning / pattern recognition techniques to classify images as either indicating a disease or not. That can be specifically useful for the detection of deadly diseases such as cancer and tuberculosis. Machines have an advantage over human expertise when it comes to identifying small changes in pixel values that are visually rather impossible to see, that may indicate the presence or absence of the disease. This project is an exercise in this direction to automate the disease detection process with achievement of high accuracy to be completed in a reasonably short amount of time, making it virtually online.

Prognosis of cancer patients can be improved through its early detection. Visually identifying cancer in its earlier stages is at times difficult, and is subject to visual interpretation by the oncologist. That introduces chances of error, which may have debilitating effect on the health of the patient. Machines can help in accurate earlier diagnosis of the cancer, and may help in proper treatment of the patient with higher chances of survival post-treatment.

Breast cancer is one of the leading causes of death among women after lung cancer [13]. Every year around 40,000 women in Pakistan die due to it [14]. Early detection may help in saving lives and enhancing the quality of life for the patients [15]. Local hospitals have solutions that take in MRI images and help the oncologists in manually identifying the disease. However, they lack in disease diagnosis. Such a solution, if built with desired accuracy, can find a ready local market.

Project Summary:

The prototype of the complete solution that we envisage at the end of three years, will take in breast mammograms as input and process them using image processing and machine learning techniques. After the analysis, the system will diagnose and identify the absence or presence of cancer. If a cancer has been identified, it finds its location, shape, and size. It would then be made as an input to the module that does visualization for the ease of the medical practitioner.

Since, it would be a complete Computer Aided Diagnostic (CAD) system (see Fig. 1), it would need to be made sufficiently efficient in doing diagnosis. That would require use of latest techniques and models. Using the cutting edge techniques and models is easier said than done, as it is not just the concept that is to be understood and implemented, but also the architecture for which the codes are written and run. Understanding the architecture to use it to the best of what it can provide also requires some research effort. Both the techniques and the architecture would be new and utilizing them in the best possible way won’t be possible to achieve without first doing background literature survey as part of a proper research methodology.

Fig. 1. Breast cancer detection: a complete cycle.

Fig. 1 gives a complete cycle of the CAD system. It takes in the mammogram, applies image processing techniques to enhance the image, which is then fed to the feature extraction techniques. Their output is given to the classifier which is then trained to classify unseen instances.

Using the cutting end technology, the cancer detection images will be combined for visualizing a 3-D rotating image for improved diagnostics, and identification of size and location of the tumor. It will be more accurate than mammography in pinpointing the size and location of cancer tumors in dense breast tissue. The 3-D rotating image will be able to zoom in, zoom out, and rotate the diagnostics image for better analysis in dense breast tissue.

Fig. 2. Visualization of the cancer on the mammograms.

Fig. 2. gives a glimpse of how visualization could be done for the cancer on the mammograms. The CAD system will not just tell if the cancer is present or not, but will also locate it, find its shape and size and render it in 3D (not shown).

 Applications Domains:

    • Brain Tumor Detection
    • Breast Cancer Detection
    • Tuberculosis Identification

Industrial Partners

Radish Medical Solutions (now Xylexa)

De Gate Pvt Ltd  NTN/STN: 7394103-0