Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one

A survey on deep learning in medical image analysis

Graphical Abstract Sample Work

Deep learning has rapidly transformed deep learning medical imaging by enabling automated extraction of complex patterns from large imaging datasets. Medical image analysis plays a critical role in disease diagnosis, treatment planning, and monitoring, yet traditional methods often rely on handcrafted features and significant expert intervention. Recently, AI in medical imaging, particularly CNN medical imaging approaches such as convolutional neural networks, has achieved strong performance in image classification, segmentation, and detection across multiple modalities. These techniques are widely applied in radiology, pathology, ophthalmology, and cardiology, improving diagnostic accuracy and clinical efficiency.To communicate these advances clearly, researchers increasingly use AI in medical imaging visualization, including medical image analysis graphical abstract formats and research graphical abstract design. Support from scientific graphical abstract service providers also enables clear medical AI illustration for academic publications. Despite progress, challenges remain in data availability, model generalizability, and clinical integration.

A collage instead of a flowchart: why it suits a survey paper

Most graphical abstracts summarise one study with a methods-to-results flow. This sample takes a different route because its source is a survey: the widely cited review by Litjens and colleagues, "A survey on deep learning in medical image analysis", published in Medical Image Analysis in 2017. A survey has no single cohort or result. Its contribution is breadth, so the figure shows breadth: a three-by-three collage of imaging applications where deep learning had achieved state-of-the-art performance, with a numbered caption that guides the reader from the top left to the bottom right.

The intended readers are radiologists, pathologists, computer scientists and clinical researchers who need to see quickly which problems the review covers.

The nine panels and what each represents

The caption lists the applications in reading order. Each panel uses a representative image from that domain:

  1. Mammographic mass classification, shown with a mammogram.
  2. Segmentation of lesions in the brain, shown with an axial brain scan.
  3. Leak detection in airway tree segmentation, shown as a three-dimensional bronchial tree.
  4. Diabetic retinopathy classification, shown as a labelled fundus illustration with haemorrhages, aneurysm, cotton wool spots and hard exudates.
  5. Prostate segmentation, with the gland highlighted in red on a pelvic image.
  6. Nodule classification, with a boxed nodule on a chest CT slice.
  7. Breast cancer metastases detection, with a histopathology section beside a heatmap of predicted tumour regions.
  8. Skin lesion classification, shown with a pigmented lesion.
  9. Bone suppression, shown with a chest radiograph.

The selection spans classification, segmentation, detection and image enhancement tasks across radiology, ophthalmology, digital pathology and dermatology.

What the review established about deep learning in imaging

Before deep learning, medical image analysis relied heavily on handcrafted features designed by experts for each task. Convolutional neural networks (CNNs) learn features directly from pixel data, and the Litjens survey documented how quickly they were adopted across modalities and organs. The applications in the collage illustrate the main task types the review discusses: assigning a label to an image or region (classification), outlining structures voxel by voxel (segmentation), locating abnormalities (detection) and improving image content (as in bone suppression on chest radiographs).

The review also discussed persistent obstacles, including limited annotated data, variation between scanners and institutions that affects generalisability, and the gap between benchmark performance and clinical integration. A graphical abstract cannot show those caveats in detail, which is one reason the text caption speaks of results in specific applications rather than claiming general clinical readiness.

Layout decisions: grid, caption and citation block

The design is deliberately restrained. Images fill the left two-thirds of the canvas edge to edge, with no borders or labels on the images themselves, so each one reads as a clean example. The right third holds a short heading, the numbered list, and a dark citation block containing the full author list, journal, volume, pages and DOI.

Keeping labels off the images has trade-offs. It preserves the visual impact of the collage, but readers must map numbers to positions. The phrase "From top-left to bottom-right" resolves that, and it is a useful convention to borrow whenever numbered panels are not individually labelled. Mixed image sources (radiographs, an illustration, photographs and microscopy) also mean that colour and contrast vary; arranging darker images away from each other helps balance the grid.

Reporting and permissions for AI imaging figures

Two practical points apply to figures like this. First, medical images reused from publications or datasets need appropriate permission or an open licence, and any patient images must be de-identified. Second, authors of original AI imaging studies are increasingly asked to follow reporting guidance such as the Checklist for Artificial Intelligence in Medical Imaging (CLAIM). A graphical abstract for such a study should state the task, modality and dataset clearly, which this survey-level collage does at the application level.

When to choose a collage for your own paper

A collage works when the message is "this approach applies across many problems", as in reviews, methods papers with multiple validation tasks, or dataset papers. It works less well for a study with one primary result, where a flow from input to output is clearer. If you adapt this format, keep the number of panels manageable, number them in reading order, and make sure each panel corresponds to a section of the paper.

Pubrica's graphical abstract service can propose either format from a manuscript. For review authors, literature review and gap analysis support helps define the scope that such a figure ultimately depicts.

Questions about this sample

Why does this graphical abstract have no arrows or flow diagram?

The source is a survey of the field rather than a single experiment, so there is no linear sequence from data to result to depict. A collage of nine applications communicates the breadth of problems covered, which is the main contribution of a review. A numbered caption replaces arrows as the guide to reading order.

Which imaging tasks are represented in the nine panels?

The panels cover classification (mammographic masses, diabetic retinopathy, lung nodules, skin lesions), segmentation (brain lesions, prostate, airway tree), detection (breast cancer metastases in histopathology) and image enhancement (bone suppression on chest radiographs). Together they span radiology, ophthalmology, pathology and dermatology.

Can published medical images be reused in a graphical abstract?

Only with appropriate permission or under a licence that allows reuse, with the source acknowledged. Patient images must also be de-identified. For original studies, authors usually draw on their own data or openly licensed datasets, which avoids permission delays at the submission stage.

pubrica services with experts

Research your Services with our experts

Delivered on-time or your money back

Give yourself the academic edge today

Each order includes

  • On-time delivery or your money back
  • A fully qualified writer in your subject
  • In-depth proofreading by our Quality Control Team
  • 100% confidentiality, the work is never re-sold or published
  • Standard 7-day amendment period
  • A paper written to the standard ordered
  • A detailed plagiarism report
  • A comprehensive quality report

Enquire Now

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.