Interview with Prof. Dr. Mrinal Mandal—Winner of the Journal of Imaging 2022 Best Paper Award

We wish to congratulate Prof. Dr. Mrinal Mandal for winning the J. Imaging 2022 Best Paper Award with his publication “Skin Lesion Segmentation Using Deep Learning with Auxiliary Task”.

Name: Prof. Dr. Mrinal Mandal
Affiliation: Digital Image Analysis Laboratory, Department of Electrical and Computer Engineering, University of Alberta, Edmonton, AB T6G 1H9, Canada
Research Interests: medical image analysis; computer-aided diagnosis; computer vision; machine learning; deep learning

The following is an interview with Prof. Dr. Mrinal Mandal:

1. Could you briefly introduce yourself to the readers? Could you introduce your current research direction and progress?

I am a tenured full professor at the University of Alberta, Canada, in the Department of Electrical and Computer Engineering. My research primarily focuses on image analysis, machine learning and computer-aided diagnosis. I collaborate with doctors at the University of Alberta Hospital for developing computer-aided diagnosis algorithms (e.g., cancer detection software). I also collaborate with industry partners for developing image analysis algorithms for industrial use.

2. Could you please briefly introduce the main content of the winning paper?

This paper proposes a novel deep learning architecture for skin lesion segmentation, which can be subsequently used for skin cancer detection. The convolutional neural network (CNN) architectures are state-of-the-art techniques for image segmentation. In typical CNN models, an image passes through many convolution layers and segmentation is achieved at the end. In our paper, we use a traditional CNN backbone in the beginning, and the backbone output then proceeds into two parallel branches. Branch-1 focuses on the contour (edge) of the lesion and branch-2 focuses on the segmentation mask. The two branches interact with each other and boost each other’s performance, resulting in an overall superior performance. For example, the edge information from branch-1 can make the mask prediction in branch-2 focus more on edges. Similarly, mask information in branch-2 contains contextual and dense pixel information and can guide the learning of the lesion edges in branch-1. Multiscale feature aggregation is also used where the edge and segmentation mask prediction are initially performed at low resolution, then at medium resolution and finally at high resolution.

3. Could you describe the difficulties and breakthrough innovations in this research field?

Machine learning and artificial intelligence techniques have shown great promise in recent years. These techniques need a large amount of data for model training. In medical image analysis, publicly available images/data, especially labelled data, are limited. This is perhaps one of the major limitations which hinders growth in medical image analysis research.

4. What appealed to you about the journal that made you want to submit your paper? How was your experience submitting to J. Imaging?

There are several reasons for choosing J. Imaging. First, J. Imaging publishes excellent quality articles. Secondly, as it is an open-source publication, it has good visibility and citations. Finally, the journal has a very fast processing time – I am not aware of any other journal in our research area which can beat it. 

5. Do you have any advice for aspiring young researchers looking to make a meaningful impact in their respective fields?

Image analysis is becoming important in many application areas. The number of researchers is growing at a rapid rate. My advice will be to conduct thorough research. A good-quality publication is likely to be more impactful than a few average-quality publications. For average-quality papers, not many researchers read them, and their impact is likely to be much less. Whatever you present, present it well and in a convincing manner.

6. As the winner of this award, is there something you would like to express or someone you wish to thank most?

It is always a good feeling when a paper receives an award. This gives us more incentive to work harder and produce better-quality manuscripts in the future. So, a big thank you to the J. Imaging award committee. I would also like to thank my co-authors, Dr. Lina Liu and Dr. Ying Tsui, for their hard work and important contributions.