For the first article to highlight in the freshly-conceived Editor’s Spotlight, from TOCHI Issue 23:1 I selected a piece of work that strongly reminded me of the context of some of my own graduate research, which took place embedded in a neurosurgery department. In my case, our research team (consisting of both physicians and computer scientists) sought to improve the care of patients who were often referred to the university hospital with debilitating neurological conditions and extremely grave diagnoses.

When really strong human-computer interaction research collides with real-world problems like this, in my experience compelling clinical impact and rigorous research results are always hard-won but in the end they are well worth the above-and-beyond efforts required to make such interdisciplinary collaborations fly.

And the following TOCHI Editor’s Spotlight paper, in my opinion, is an outstanding example of such a contribution.

IN THE SPOTLIGHT:

Navigating Giga-pixel Images in Digital Pathology

The diagnosis of cancer is serious business, yet in routine clinical practice pathologists still work on microscopes, with physical slides, because digital pathology runs up against many barriers—not the least of which are the navigational challenges raised by panning and zooming through huge (and I mean huge) image datasets on the order of multiple gigapixels. And that’s just for a single slide.

Few illustrations grace the article, but those that do—

They stop the reader cold.

Extract from a GI biopsy, showing malignant tissue at 400x magnification. (Fig. 3)

The ruddy and well-formed cells of healthy tissue from a GI biopsy slowly give way to an ill-defined frontier of pathology, an ever-expanding redoubt for the malignant tissue lurking deep within. One cannot help but be struck by the subtext that these images represent the lives of patients that face a dire health crisis.

Only by finding, comparing, and contrasting this tissue to other cross-sections and slides—scanned at 400x magnification and a startling 100,000 dots per inch—can the pathologist arrive at a correct and accurate diagnosis as to the type and extent of the malignancy.

This article stands out because it puts into practice—and challenges—accepted design principles for the navigation of such gigapixel images, against the backdrop of real work by medical experts.

These are not laboratory studies that strive for some artificial measure of “ecological validity”—no, here the analyses take place in the context of the real work of pathologists (using archival cases) and yet the experimental evaluations are still rigorous and insightful. There is absolutely no question of validity and the stakes are clearly very high.

While the article focuses on digital pathology, the insights and perspectives it raises (not to mention the interesting image navigation and comparison tasks motivated by clinical needs) should inform, direct, and inspire many other efforts to improve interfaces for navigation through large visualizations and scientific data-sets.

 


Roy Ruddle, Thomas Rhys, Rebecca Randell, Phil Quirke, and Darren Treanor. 2016. The Design and Evaluation of Interfaces for Navigating Gigapixel Images in Digital Pathology. ACM Trans. Comput.-Hum. Interact. 23, 1, Article 5 (February 2015), 29 pages. DOI= http://dx.doi.org/10.1145/2834117