Distance measures for image segmentation evaluation

The task considered in this paper is performance evaluation of region segmentation algorithms in the ground-truth-based paradigm. Given a machine segmentation and a ground-truth segmentation, performance measures are needed. We propose to consider the image segmentation problem as one of data clustering and, as a consequence, to use measures for comparing clusterings developed in statistics and machine learning. By doing so, we obtain a variety of performance measures which have not been used before in image processing. In particular, some of these measures have the highly desired property of being a metric. Experimental results are reported on both synthetic and real data to validate the measures and compare them with others.

[Verification of normative values for the Swiss version of the Freiburg speech intelligibility test.]

BACKGROUND AND OBJECTIVE: In the Swiss version of the Freiburg speech intelligibility test five test words from the original German recording which are rarely used in Switzerland have been exchanged. Furthermore, differences in the transfer functions between headphone and loudspeaker presentation are not taken into account during calibration. New settings for the levels of the individual test words in the recommended recording and small changes in calibration procedures led us to make a verification of the currently used normative values.PATIENTS AND METHODS: Speech intelligibility was measured in 20 subjects with normal hearing using monosyllabic words and numbers via headphones and loudspeakers.RESULTS: On average, 50% speech intelligibility was reached at levels which were 7.5 dB lower under free-field conditions than for headphone presentation. The average difference between numbers and monosyllabic words was found to be 9.6 dB, which is considerably lower than the 14 dB of the current normative curves.CONCLUSIONS: There is a good agreement between our measurements and the normative values for tests using monosyllabic words and headphones, but not for numbers or free-field measurements.

Early assessment of brain maturation by MR imaging segmentation in neonates and premature infants

PURPOSE: We evaluated the impact of premature extrauterine life on brain maturation. PATIENTS AND METHODS: Twelve neonates underwent MR imaging at 40 (39.64 +/- 0.98) weeks (full term). Fifteen premature infants underwent 2 MR imaging examinations, after birth (preterm at birth) and at 40 weeks (41.03 +/- 1.33) (preterm at term). A 3D MR imaging technique was used to measure brain volumes compared with intracranial volume: total brain volume, cortical gray matter, myelinated white matter, unmyelinated white matter, basal ganglia (BG), and CSF. RESULTS: The average absolute volume of intracranial volume (269.8 mL +/- 36.5), total brain volume (246.5 +/- 32.3), cortical gray matter (85.53 mL +/- 22.23), unmyelinated white matter (142.4 mL +/-14.98), and myelinated white matter (6.099 mL +/-1.82) for preterm at birth was significantly lower compared with that for the preterm at term: the average global volume of intracranial volume (431.7 +/- 69.98), total brain volume (391 +/- 66,1), cortical gray matter (179 mL +/- 41.54), unmyelinated white matter (185.3 mL +/- 30.8), and myelinated white matter (10.66 mL +/- 3.05). It was also lower compared with that of full-term infants: intracranial volume (427.4 mL +/- 53.84), total brain volume (394 +/- 49.22), cortical gray matter (181.4 +/- 29.27), unmyelinated white matter (183.4 +/- 27.37), and myelinated white matter (10.72 +/- 4.63). The relative volume of cortical gray matter (30.62 +/- 5.13) and of unmyelinated white matter (53.15 +/- 4.8) for preterm at birth was significantly different compared with the relative volume of cortical gray matter (41.05 +/- 5.44) and of unmyelinated white matter (43.22 +/- 5.11) for the preterm at term. Premature infants had similar brain tissue volumes at 40 weeks to full-term infants. CONCLUSION: MR segmentation techniques demonstrate that cortical neonatal maturation in moderately premature infants at term and term-born infants was similar.

Affective and linguistic processing of speech prosody: DC potential studies

Speech melody or prosody subserves linguistic, emotional, and pragmatic functions in speech communication. Prosodic perception is based on the decoding of acoustic cues with a predominant function of frequency-related information perceived as speaker's pitch. Evaluation of prosodic meaning is a cognitive function implemented in cortical and subcortical networks that generate continuously updated affective or linguistic speaker impressions. Various brain-imaging methods allow delineation of neural structures involved in prosody processing. In contrast to functional magnetic resonance imaging techniques, DC (direct current, slow) components of the EEG directly measure cortical activation without temporal delay. Activation patterns obtained with this method are highly task specific and intraindividually reproducible. Studies presented here investigated the topography of prosodic stimulus processing in dependence on acoustic stimulus structure and linguistic or affective task demands, respectively. Data obtained from measuring DC potentials demonstrated that the right hemisphere has a predominant role in processing emotions from the tone of voice, irrespective of emotional valence. However, right hemisphere involvement is modulated by diverse speech and language-related conditions that are associated with a left hemisphere participation in prosody processing. The degree of left hemisphere involvement depends on several factors such as (i) articulatory demands on the perceiver of prosody (possibly, also the poser), (ii) a relative left hemisphere specialization in processing temporal cues mediating prosodic meaning, and (iii) the propensity of prosody to act on the segment level in order to modulate word or sentence meaning. The specific role of top-down effects in terms of either linguistically or affectively oriented attention on lateralization of stimulus processing is not clear and requires further investigations.

Speech understanding in quiet and in noise with the bone-anchored hearing aids Baha Compact and Baha Divino

CONCLUSIONS: Speech understanding is better with the Baha Divino than with the Baha Compact in competing noise from the rear. No difference was found for speech understanding in quiet. Subjectively, overall sound quality and speech understanding were rated better for the Baha Divino. OBJECTIVES: To compare speech understanding in quiet and in noise and subjective ratings for two different bone-anchored hearing aids: the recently developed Baha Divino and the Baha Compact. PATIENTS AND METHODS: Seven adults with bilateral conductive or mixed hearing losses who were users of a bone-anchored hearing aid were tested with the Baha Compact in quiet and in noise. Tests were repeated after 3 months of use with the Baha Divino. RESULTS: There was no significant difference between the two types of Baha for speech understanding in quiet when tested with German numbers and monosyllabic words at presentation levels between 50 and 80 dB. For speech understanding in noise, an advantage of 2.3 dB for the Baha Divino vs the Baha Compact was found, if noise was emitted from a loudspeaker to the rear of the listener and the directional microphone noise reduction system was activated. Subjectively, the Baha Divino was rated statistically significantly better in terms of overall sound quality.

Comparison of musical activities of cochlear implant users with different speech-coding strategies

Speech coding might have an impact on music perception of cochlear implant users. This questionnaire study compares the musical activities and perception of postlingually deafened cochlear implant users with three different coding strategies (CIS, ACE, SPEAK) using the Munich Music Questionnaire. Overall, the self-reported perception of music of CIS, SPEAK, and ACE users did not differ by very much.

Clinical validation of an automated vessel-segmentation software of the extracranial-carotid arteries based on 3D-MRA: A prospective study

OBJECTIVES: To determine the accuracy of automated vessel-segmentation software for vessel-diameter measurements based on three-dimensional contrast-enhanced magnetic resonance angiography (3D-MRA). METHOD: In 10 patients with high-grade carotid stenosis, automated measurements of both carotid arteries were obtained with 3D-MRA by two independent investigators and compared with manual measurements obtained by digital subtraction angiography (DSA) and 2D maximum-intensity projection (2D-MIP) based on MRA and duplex ultrasonography (US). In 42 patients undergoing carotid endarterectomy (CEA), intraoperative measurements (IOP) were compared with postoperative 3D-MRA and US. RESULTS: Mean interoperator variability was 8% for measurements by DSA and 11% by 2D-MIP, but there was no interoperator variability with the automated 3D-MRA analysis. Good correlations were found between DSA (standard of reference), manual 2D-MIP (rP=0.6) and automated 3D-MRA (rP=0.8). Excellent correlations were found between IOP, 3D-MRA (rP=0.93) and US (rP=0.83). CONCLUSION: Automated 3D-MRA-based vessel segmentation and quantification result in accurate measurements of extracerebral-vessel dimensions.

Formal speech stylistics and type A behavior in 38 subjects during nonstress interviews

Open-ended interviews of 90 min length of 38 patients were analyzed with respect to speech stylistics, shown by Schucker and Jacobs to differentiate individuals with type A personality features from those with type B. In our patients, Type A/B had been assessed by the Bortner Personality Inventory. The stylistics studied were: repeated words swallowed words, interruptions, simultaneous speech, silence latency (between question and answer) (SL), speed of speech, uneven speed of speech (USS), explosive words (PW), uneven speech volume (USV), and speech volume. Correlations between both raters for all speech categories were high. Positive correlations between extent of type A and SL (r = 0.33; p = 0.022), USS (r = 0.51; p = 0.002), PW (r = 0.46; p = 0.003) and USV (r = 0.39; p = 0.012) were found. Our results indicate that the speech in nonstress open-ended interviews of type A individuals tends to show a higher emotional tension (positive correlations for USS PW and USV) and is more controlled in conversation (positive correlation for SL).

Semi-automatic segmentation of fetal cardiac cavities: progress towards an automated fetal echocardiogram

OBJECTIVE: To develop a novel application of a tool for semi-automatic volume segmentation and adapt it for analysis of fetal cardiac cavities and vessels from heart volume datasets. METHODS: We studied retrospectively virtual cardiac volume cycles obtained with spatiotemporal image correlation (STIC) from six fetuses with postnatally confirmed diagnoses: four with normal hearts between 19 and 29 completed gestational weeks, one with d-transposition of the great arteries and one with hypoplastic left heart syndrome. The volumes were analyzed offline using a commercially available segmentation algorithm designed for ovarian folliculometry. Using this software, individual 'cavities' in a static volume are selected and assigned individual colors in cross-sections and in 3D-rendered views, and their dimensions (diameters and volumes) can be calculated. RESULTS: Individual segments of fetal cardiac cavities could be separated, adjacent segments merged and the resulting electronic casts studied in their spatial context. Volume measurements could also be performed. Exemplary images and interactive videoclips showing the segmented digital casts were generated. CONCLUSION: The approach presented here is an important step towards an automated fetal volume echocardiogram. It has the potential both to help in obtaining a correct structural diagnosis, and to generate exemplary visual displays of cardiac anatomy in normal and structurally abnormal cases for consultation and teaching.