25799

ارزيابي رمزگذاري ويژگي هاي بينايي در نمايش همزمان چند ويژگي

كارشناسي ارشد

مهندسي پزشكي

1397

1400/08/29

دكتر محمدرضا دليري

مهندسي برق

تصوير يك شي معمولاً از ويژگي¬هاي مختلفي تشكيل شده است (به عنوان مثال، رنگ، جهت گيري و حركت)، كه توسط مسيرهاي بصري تفكيك شده پردازش مي¬شوند و توسط مناطق ويژه حسي و عملكرد مغز رمزگذاري مي¬شوند. با اين حال، ما يك شي را به عنوان يك مفهوم يكپارچه درك مي-كنيم. چگونگي ادغام اين ويژگي¬هاي جدا شده و دستيابي به درك دقيق از اشيا يك چالش اساسي براي محققان حوزه¬ي بينايي است. ما در مطالعه رفتاري با سوژه انساني نشان داديم كه رنگ با اختلاف معنا¬داري سريع¬تر از اورينتيشن درك مي¬شود(ويلكاكسون P<0.01). با مقايسه زمان عكس العمل بايندينگ مشاهده كرديم كه علارغم رمزگذاري شدن سريع¬تر رنگ نسبت به اورينتيشن زمان عكس¬العمل اورينتيشن سريع تر از رنگ مي-باشد(ويلكاكسون P<0.01). در حالت بايندينگ هم اين اختلاف بين دو ويژگي ديده شده است. كه مي¬توان اين فرضيه را مطرح كرد كه در حالات بايدينگ، تك ويژگي¬ها به صورت توالي بررسي مي-شوند و اين اختلاف زمان عكس¬العمل بين بايندينگ و تك ويژگي زمان مورد نياز براي بايند كردن ويژگي¬ها مي¬باشد. ما همچنين به ارزيابي رمزگذاري رنگ و جهت در ديتاي نوروني در ناحيه بين آهيانه¬اي جانبي پرداختيم. مطالعات متعددي نشان دادند كه توجه بر اساس ويژگي به نورون¬هاي LIP اجازه مي¬دهد تا چندين ويژگي بينايي را ادغام كند. LIP با هر دو مسير بصري قدامي و خلفي در ارتباط است و مي تواند ويژگي هاي بصري مانند جهت، رنگ و فرم را رمزگذاري كند. در ابتدا با استفاده از آناليز منحني مشخصه عملكرد نشان داديم كه جهت با اختلاف معناداري نسبت به رنگ زودتر و همچنين با قدرت بيشتري رمزگذاري مي¬شود (ويلكاكسون p<0.05، اختلاف زماني 46 ميلي¬ثانيه). ما اين مشاهدات را با استفاده از روش رمزگشايي جمعيتي با استفاده از طبقه بند ماشين بردار پشتيبان تأييد كرديم. در انتها نشان داديم كه رمزگذاري هر ويژگي با اختلاف معناداري نسبت به مچينگ همان ويژگي اتفاق مي¬افتد(آزمون t p<0.01).

1400/10/11

Evaluate the encoding of visual features in the simultaneous display of multiple features

11/20/2022 12:00:00 AM

Objects in our visual world usually have several features dimensions simultaneously (such as color, direction, and orientation). However, previous studies claim that our brain integrates these feature and cause us to perceive these object as a whole, regardless of that different features are encoded in a separate area and visual pathways. Integrating these features separately and precisely from objects is a fundamental challenge for researchers in the field of vision, which is referred to as the Binding problem. Much research has been done on how to bind the features observed in perceiving the environment. They concluded that the integration of features exists in many perceptual processes. We designed a behavioral task to investigate this by comparing the timing of features’ perception in either a single-feature experiment or an integration experiment. In the first experiment, the subject was asked only to attend to one feature dimension (color/orientation) and report it. In the second experiment, the subject was required to attend to both features simultaneously and report as a single object, including one color and orientation. Also, we considered different target’ exposure times to realize the minimum time required to perceive these features. Our result demonstrates that subjects attend and respond to the objects’ features sequentially. Furthermore, perception time is different from feature processing time. Because the color feature was perceived earlier than the orientation, but the reaction time for the color was longer than the orientation. Moreover, this sequence was observed both in perception and reaction time in the integration experiment, similar to the single-feature experiment. Besides, due to sequential feature perception and subjects’ response, we claim that visual features integration in the simultaneous display is a post-perception process. We also examined color and direction encoding in neuronal data in the lateral intraparietal area (LIP) region. Numerous studies have been performed on the role of the LIP region in spatial visual functions (e.g., spatial attention, saccadic eye movements), and it has been shown that feature-based attention allows LIP neurons to integrate multiple visual features representations from upstream areas flexibly. LIP is interconnected with both dorsal and ventral stream visual areas and can encode visual features such as color, direction, and form, particularly when stimuli are task-relevant. Initially, using Receiver operating characteristic (ROC) showed direction encoded significantly earlier and with greater power than color. We also used the direction and color of the preferred and anti-preferred neurons to examine the coding of these features. The results showed a sequence similar to the case in which the features with the most differences were selected. We validated these observations using the population decoding method in which we trained the Support vector machines (SVMs) classifier for color and direction decoding. Finally, we examined the matching encoding between two identical features and between two different features and showed that the decoding of each feature occurred with a significant difference earlier than the matching of the same feature.

رمزگذاري شدن ويژگي , ويژگي¬هاي چندگانه , درك يكپارچه , داده¬¬هاي رفتاري , بايندينگ ويژگي¬ها

Visual Features Encoding , Multiple Features , Integrated Perception , Behavioral Data , Features integration