Medical image processing applications typically demand highly accurate image segmentation. However, existing segmentation approaches exhibit performance degradation when faced with diverse medical imaging modalities and varied segmentation target sizes. In this paper, we propose and evaluate a dilated One-to-Many U-Net deep learning model that addresses these challenges. The proposed model comprises of four rows of encoder-decoder modules, with each module consisting of three trainable blocks with different layers. The last three rows of the U-Net are extended versions of the three blocks in the first row, with the encoder-decoder blocks connected through the skip connections to the previous rows. The outputs of the last blocks from the last three rows in the decoder are concatenated, and finally, a dilation network is employed to improve the small target segmentation in different medical images. Two datasets have been used for the evaluation: the HC18 grand challenge ultrasound dataset for fetal head segmentation and the Multi-site MRI dataset, including the BIDMC and HK sites, for prostate segmentation in MRI images. The proposed approach achieved Dice and Jaccard coefficients of 96.54% and 93.93%, respectively, for the HC18 grand challenge dataset, 96.76% and 93.97% for the BIDMC site dataset, and 92.58% and 86.96% for the HK site dataset. Statistical analyses showed that the proposed model outperformed several other U-Net-based models.

In this study, we proposed a novel automated speech quality estimation model capable of evaluating perceptual dysphonia severity and breathiness in audio samples, ensuring alignment with expert-rated assessments. The proposed model integrates Whisper ASR embeddings with Mel spectrograms augmented by second-order delta features combined with a sequential-attention fusion network feature mapping path. This hybrid approach enhances the model’s sensitivity to phonetic, high level feature representation and spectral variations, enabling more accurate predictions of perceptual speech quality. A sequential-attention fusion network feature mapping module captures long-range de-pendencies through the multi-head attention network, while LSTM layers refine the learned representations by modeling temporal dynamics. Comparative analysis against state-of-the-art methods for dysphonia assessment demonstrates our model’s superior generalization across test samples. Our findings underscore the effectiveness of ASR-derived embeddings alongside the deep feature mapping structure in speech quality assessment, offering a promising pathway for advancing automated evaluation systems.

In this study, we propose a novel automated model for speech quality estimation that objectively evaluates perceptual dysphonia severity and breathiness in audio samples, demonstrating strong correlation with expert ratings. The proposed model integrates Whisper encoder embeddings with Mel spectrograms augmented by second-order delta features combined with a sequential-attention fusion network feature mapping path. This hybrid approach enhances the model’s sensitivity to phonetic, high-level feature representation, and spectral variations, enabling more accurate predictions of perceptual speech quality. A sequential-attention fusion network feature mapping module captures long-range dependencies through the multi-head attention network, while LSTM layers refine the learned representations by modeling temporal dynamics. Comparative analysis against state-of-the-art methods for dysphonia assessment demonstrates our model’s better correlation with clinician’s judgments across test samples. Our findings underscore the effectiveness of ASR-derived embeddings alongside the deep feature mapping structure in disordered speech quality assessment, offering a promising pathway for advancing automated evaluation systems.

Purpose: Best practices in audiology recommend assessing speech understanding in noisy environments, especially for those with communication difficulties. Speech-in-noise (SiN) assessments such as the QuickSIN are used for validating signal processing in hearing aids (HAs) and are linked to HA satisfaction. This project seeks to enhance QuickSIN test efficiency by applying recent advancements in automatic speech recognition (ASR) technologies. Method: Twenty-three adults with sensorineural hearing loss were fitted bilaterally with Unitron Moxi HAs and were administered the QuickSIN test in low and high reverberation environments. Testing was performed with two different HA programs: an omnidirectional program and a fixed directional microphone program. QuickSIN sentences were presented from 0° azimuth and competing babble from either 0°, laterally from 90° or 270°, or simultaneously from 90°, 180°, and 270° azimuths. Participants’ verbal responses to QuickSIN stimuli were scored by an audiologist and were recorded in parallel for offline transcription and scoring by ASR models from Amazon, Microsoft, NVIDIA, and Picovoice. The ASR-derived QuickSIN scores were compared to the corresponding audiologist-derived scores. Results: Repeated Measures ANOVA results revealed that all ASR models overestimated the QuickSIN scores across most test conditions. Bland-Altman analyses showed that the Amazon ASR model had the least bias and the narrowest range for the limits of agreement, in comparison to the manual scoring by an experienced audiologist. Conclusions: Some ASR models, such as Amazon, demonstrated performance comparable to that of an audiologist in automatically scoring QuickSIN tests. However, further refinements are necessary to increase the robustness of the ASR models in scoring low SNR loss test conditions.