-
RACR-MIL: Weakly Supervised Skin Cancer Grading using Rank-Aware Contextual Reasoning on Whole Slide Images
Authors:
Anirudh Choudhary,
Angelina Hwang,
Jacob Kechter,
Krishnakant Saboo,
Blake Bordeaux,
Puneet Bhullar,
Nneka Comfere,
David DiCaudo,
Steven Nelson,
Emma Johnson,
Leah Swanson,
Dennis Murphree,
Aaron Mangold,
Ravishankar K. Iyer
Abstract:
Cutaneous squamous cell cancer (cSCC) is the second most common skin cancer in the US. It is diagnosed by manual multi-class tumor grading using a tissue whole slide image (WSI), which is subjective and suffers from inter-pathologist variability. We propose an automated weakly-supervised grading approach for cSCC WSIs that is trained using WSI-level grade and does not require fine-grained tumor an…
▽ More
Cutaneous squamous cell cancer (cSCC) is the second most common skin cancer in the US. It is diagnosed by manual multi-class tumor grading using a tissue whole slide image (WSI), which is subjective and suffers from inter-pathologist variability. We propose an automated weakly-supervised grading approach for cSCC WSIs that is trained using WSI-level grade and does not require fine-grained tumor annotations. The proposed model, RACR-MIL, transforms each WSI into a bag of tiled patches and leverages attention-based multiple-instance learning to assign a WSI-level grade. We propose three key innovations to address general as well as cSCC-specific challenges in tumor grading. First, we leverage spatial and semantic proximity to define a WSI graph that encodes both local and non-local dependencies between tumor regions and leverage graph attention convolution to derive contextual patch features. Second, we introduce a novel ordinal ranking constraint on the patch attention network to ensure that higher-grade tumor regions are assigned higher attention. Third, we use tumor depth as an auxiliary task to improve grade classification in a multitask learning framework. RACR-MIL achieves 2-9% improvement in grade classification over existing weakly-supervised approaches on a dataset of 718 cSCC tissue images and localizes the tumor better. The model achieves 5-20% higher accuracy in difficult-to-classify high-risk grade classes and is robust to class imbalance.
△ Less
Submitted 29 August, 2023;
originally announced August 2023.
-
Leveraging Cluster Analysis to Understand Educational Game Player Experiences and Support Design
Authors:
Luke Swanson,
David Gagnon,
Jennifer Scianna,
John McCloskey,
Nicholas Spevacek,
Stefan Slater,
Erik Harpstead
Abstract:
The ability for an educational game designer to understand their audience's play styles and resulting experience is an essential tool for improving their game's design. As a game is subjected to large-scale player testing, the designers require inexpensive, automated methods for categorizing patterns of player-game interactions. In this paper we present a simple, reusable process using best practi…
▽ More
The ability for an educational game designer to understand their audience's play styles and resulting experience is an essential tool for improving their game's design. As a game is subjected to large-scale player testing, the designers require inexpensive, automated methods for categorizing patterns of player-game interactions. In this paper we present a simple, reusable process using best practices for data clustering, feasible for use within a small educational game studio. We utilize the method to analyze a real-time strategy game, processing game telemetry data to determine categories of players based on their in-game actions, the feedback they received, and their progress through the game. An interpretive analysis of these clusters results in actionable insights for the game's designers.
△ Less
Submitted 18 October, 2022;
originally announced October 2022.
-
Exploring players' experience of humor and snark in a grade 3-6 history practices game
Authors:
David J. Gagnon,
Ryan S. Baker,
Sarah Gagnon,
Luke Swanson,
Nick Spevacek,
Juliana Andres,
Erik Harpstead,
Jennifer Scianna,
Stefan Slater,
Maria O. C. Z. San Pedro
Abstract:
In this paper we use an existing history learning game with an active audience as a research platform for exploring how humor and "snarkiness" in the dialog script affect students' progression and attitudes about the game. We conducted a 2x2 randomized experiment with 11,804 anonymous 3rd-6th grade students. Using one-way ANOVA and Kruskall-Wallis tests, we find that changes to the script produced…
▽ More
In this paper we use an existing history learning game with an active audience as a research platform for exploring how humor and "snarkiness" in the dialog script affect students' progression and attitudes about the game. We conducted a 2x2 randomized experiment with 11,804 anonymous 3rd-6th grade students. Using one-way ANOVA and Kruskall-Wallis tests, we find that changes to the script produced measurable results in the self-reported perceived humor of the game and the likeability of the player character. Different scripts did not produce significant differences in player completion of the game, or how much of the game was played. Perceived humor and enjoyment of the game and its main character contributed significantly to progress in the game, as did self-perceived reading skill.
△ Less
Submitted 18 October, 2022;
originally announced October 2022.
-
A Pilot Study on Teacher-Facing Real-Time Classroom Game Dashboards
Authors:
Luke Swanson,
David Gagnon,
Jennifer Scianna
Abstract:
Educational games are an increasingly popular teaching tool in modern classrooms. However, the development of complementary tools for teachers facilitating classroom gameplay is lacking. We present the results of a participatory design process for a teacher-facing, real-time game data dashboard. This two-phase process included a workshop to elicit teachers' requirements for such a tool, and a pilo…
▽ More
Educational games are an increasingly popular teaching tool in modern classrooms. However, the development of complementary tools for teachers facilitating classroom gameplay is lacking. We present the results of a participatory design process for a teacher-facing, real-time game data dashboard. This two-phase process included a workshop to elicit teachers' requirements for such a tool, and a pilot study of our dashboard prototype. We analyze post-gameplay survey and interview data to understand teachers' experiences with the tool in terms of evidence of co-design, feasibility, and effectiveness. Our results indicate the participatory design yielded a tool both useful for and usable by teachers within the context of a real class gameplay session. We advocate for the continued development of data-driven teacher tools to improve the effectiveness of games deployed in the classroom.
△ Less
Submitted 17 October, 2022;
originally announced October 2022.
