Primary care visits are higher amongst adults who report sexual dysfunction concerns after cancer treatment.

Emotion recognition plays an increasingly important role in affective computing applications, from human-computer interaction to mental health monitoring. Among various emotion recognition tasks, understanding dynamic emotional processes represents a key challenge for achieving naturalistic emotional intelligence. Emotion recognition traditionally aims to enable machines to classify discrete emotional states from physiological and behavioral signals. Similar to other machine learning problems, high-quality datasets are prerequisite for advancing this field. Consequently, several datasets have been proposed to promote emotion recognition research from multimodal signals. For example, the DEAP dataset collected EEG, GSR, blood volume pressure, and other physiological signals from 32 participants viewing music videos. MAHNOB-HCI constructed a multimodal database consisting of facial videos, audio signals, eye gaze data, and physiological signals including EEG, ECG, and GSR. Recently, K-EmoCon created a dataset for emotion recognition in naturalistic conversations, while AMIGOS, DREAMER, and other datasets have contributed additional resources focusing on discrete emotion classification in laboratory settings. Although these existing datasets greatly promote emotion recognition research, they share a fundamental limitation of discrete treatment of emotions. They treat emotional experiences as discrete, independent events rather than components of continuous emotional flow. Effectively, their emotional elicitation paradigm is in the sense that it is defined as isolated clips followed by single post-stimulus rating or continuous external annotation. Although current datasets focus on static emotion classification from isolated stimulus-response pairs, the real-world emotional experiences involve dynamic transitions between different emotional states. While recent advances in affective computing have achieved substantial performance on static emotion classification tasks, with multimodal fusion approaches reaching 75–95% accuracy on established benchmarks, this success masks a critical limitation. In natural settings, emotional states evolve over time, transitions between emotions create unique physiological signatures, and the temporal context of previous emotional states influences current responses. This static emotion paradigm overlooks the temporal nature of emotional processes and creates a significant gap between laboratory performance and real-world emotion recognition applications. Based on this limitation, we adapt the concept of emotional journey from affective gaming research, defined as a temporally ordered sequence of distinct emotional states experienced by an individual. While previous emotional journey implementations focused on adaptive gaming narratives responding to real-time emotional states, our approach systematically guides participants through predetermined emotional transitions for scientific analysis. We define our concept of as a temporally ordered sequence of emotional blocks – each block consisting of pre-stimulus emotional rating, video stimulus, and post-stimulus emotional rating, interspersed with structured inter-stimulus intervals (comprising pre-stimulus rating, neutral washout video, and post-stimulus rating) to ensure complete autonomic recovery between emotional states. This controlled paradigm enables systematic study of emotional transitions and their physiological signatures, distinct from the adaptive entertainment applications of previous emotional journey concepts. Thus, the concept of emotional journeys necessitates need for affective datasets that explicitly invoke emotions in a chronologically successive manner. However, analysis of major emotion datasets reveals systematic methodological limitations that constrain research into emotional dynamics as shown in Table  below.